Urolithin A improves Alzheimer's disease cognition and restores mitophagy and lysosomal functions

Abstract

BACKGROUND

Compromised autophagy, including impaired mitophagy and lysosomal function, plays pivotal roles in Alzheimer's disease (AD). Urolithin A (UA) is a gut microbial metabolite of ellagic acid that stimulates mitophagy. The effects of UA's long‐term treatment of AD and mechanisms of action are unknown.

METHODS

We addressed these questions in three mouse models of AD with behavioral, electrophysiological, biochemical, and bioinformatic approaches.

RESULTS

Long‐term UA treatment significantly improved learning, memory, and olfactory function in different AD transgenic mice. UA also reduced amyloid beta (Aβ) and tau pathologies and enhanced long‐term potentiation. UA induced mitophagy via increasing lysosomal functions. UA improved cellular lysosomal function and normalized lysosomal cathepsins, primarily cathepsin Z, to restore lysosomal function in AD, indicating the critical role of cathepsins in UA‐induced therapeutic effects on AD.

CONCLUSIONS

Our study highlights the importance of lysosomal dysfunction in AD etiology and points to the high translational potential of UA.

Highlights

• Long‐term urolithin A (UA) treatment improved learning, memory, and olfactory function in Alzheimer's disease (AD) mice.

• UA restored lysosomal functions in part by regulating cathepsin Z (Ctsz) protein.

• UA modulates immune responses and AD‐specific pathophysiological pathways.

1. INTRODUCTION

The prevalence of Alzheimer's disease (AD) is estimated to triple worldwide by 2050, a massive burden in our aging society. ¹ , ² Clinical manifestations include cognitive impairments and abnormal behaviors. The main pathological features of AD are amyloid plaques formed by amyloid beta (Aβ) and neurofibrillary tangles formed by phosphorylated tau. ³ Aging is a major risk factor for neurodegenerative diseases, and many hallmarks of aging play essential roles in the pathogenesis and development of AD, including compromised autophagy, ⁴ DNA damage, neuroinflammation, cellular senescence, and mitochondrial dysfunction. ⁵ , ⁶ , ⁷ Autophagy is markedly impaired in AD, and autolysosome acidification in AD mouse models induces autophagic buildup of Aβ in neurons, yielding senile plaques; thus, compromised lysosomal function is considered a driver of AD. ⁴

Mitophagy clears damaged or superfluous mitochondria. The timely removal of damaged mitochondria plays an essential role in maintaining normal physiological function and is vital for the survival and health of neurons. ⁸ , ⁹ We and others have shown that mitophagy levels in the brains of patients with AD are compromised and that mitophagy in transgenic mice, nematode models of AD, and induced Pluripotent Stem Cell (iPSC)‐derived neurons of AD patients are impaired. ¹⁰ , ¹¹ Mitophagy induction by genetic strategies or compounds improves the cognitive function of AD mice and reduces Aβ plaques and phosphorylated tau (p‐tau) in brains. ¹¹ , ¹² Many natural mitophagy inducers, including urolithin A (UA), nicotinamide riboside (NR), kaempferol, and rhapontigenin, ¹¹ , ¹³ , ¹⁴ , ¹⁵ reduce AD symptoms.

UA is a natural compound produced by gut bacteria ingesting ellagitannins and ellagic acid, a polyphenol found in pomegranates, berries, and nuts. ¹⁶ , ¹⁷ UA was discovered 40 years ago ¹⁸ and is the most conserved and studied urolithin across species, ¹⁶ but only recently has its impact on aging and diseases been explored. UA prolongs the lifespan of Caenorhabditis elegans and safeguards against physiological decline, as illustrated by improved muscle function in young animals and the prevention of age‐related muscle decline in old mice. ¹⁹ UA also enhances cellular health by increasing mitophagy and mitochondrial function and reducing detrimental inflammation. ¹¹ , ¹⁹ , ²⁰ We recently reported that short‐term (2 months) UA treatment induced mitophagy in the AD mouse brain and nematodes, and improved learning and memory in APP/PS1 (Amyloid precursor protein / Presenilin‐1) and 3xTgAD (contains APP, PS1 and MAPT mutations) mice. ¹¹ Our results were later validated. ²¹ UA treatment had anti‐neuroinflammatory effects in activated microglia, supporting the potential neuroprotective role of UA in AD brains. ²² Other studies have confirmed the anti‐inflammatory properties of UA in vivo and in vitro. ¹⁷ , ²¹

Other mechanisms of action have been proposed for UA, including the activation of the Ahr/Nrf2 pathway and its downstream antioxidative stress response, the removal Aβ from neurons, the inhibition of DYRK1A activity, ²³ and the inhibition of regulators of cancer cell proliferation. ¹⁵ , ²³ , ²⁴ UA has so far only been investigated in humans for its benefits on mitochondrial and muscle health. ¹⁷ , ²⁵ A phase I clinical study confirmed that UA was safe in healthy, sedentary older adults, and that activation of mitochondrial biomarkers in muscle and plasma was observed, consistent with previous work shown in cells and in vivo in model organisms. ²⁵ UA is well tolerated and may play a therapeutic role in the brain as it crosses the blood–brain barrier. ²⁶ , ²⁷ Collectively, growing evidence supports that UA effectively targets AD‐related neurodegeneration and holds potential as an intervention for this disease.

Previously we reported on the benefits of UA treatment of AD mice for a brief period (2 months). ¹¹ However, most AD patients are likely to receive treatment and care for an extended period. Thus, here we have used multiple AD mouse models, APP/PS1, 3xTgAD (AD), and 3xTgAD/Polβ^+/− (ADP) mice (DNA repair–deficient AD mice made by us), and each was treated with UA for 5 months to study the effects of long term UA treatment. This study strengthens support for the potential application of UA against AD.

2. METHODS

2.1. Mice

All animal experiments were performed and approved by the National Institute on Aging (NIA) Animal Care and Use Committee and met all relevant ethical regulations (study protocol number 361‐ODS‐2023). All animals were maintained at the NIA under standard conditions and fed standard animal feed. Male APP/PS1 mice and their wild‐type (WT) littermates were used for the experiments. The APP/PS1 mouse strain (stock no. 004462; The Jackson Laboratory) was obtained from Dr. Mark Mattson's laboratory.

APP/PS1 mice were treated with UA (200 mg/kg/day, Molbase) by gavage starting at 2 months and ending at 7 months when behavioral and molecular endpoints were assessed. 3xTgAD and 3xTgAD/Polβ^+/− mouse strains were generated as described previously. ²⁸ Twelve‐month‐old 3xTgAD mice and 3xTgAD/Polβ^+/− mice were treated with UA (200 mg/kg/day) by oral gavage for 5 months, ending at 17 months, followed by behavioral and molecular endpoints evaluation. Both male and female 3xTgAD and 3xTgAD/Polβ^+/− mice were used in the analyses. Statistical methods were not used to predetermine sample sizes, but our sample sizes are similar to those reported in previous studies.

2.2. Cell culture

HMC3 human microglia cells were purchased from ATCC. HMC3 was cultured in Eagle's Minimum Essential Medium (EMEM; Gibco) with 10% fetal bovine serum (FBS) in a humidified incubator with 5% CO₂ at 37°C. HEK293 APP‐Swedish cells were maintained in high‐glucose DMEM (Gibco) supplemented with 10% FBS (Gibco, #30084) and cultured in a humidified atmosphere with 5% CO₂ at 37°C. When HMC3 cells reached 60% confluence, cells were transfected with (20 nM) siRNAs (KeyGEN BioTECH), and 4 h later were co‐cultured with 10 μM Aβ₄₂, 30 μM UA, 3 μM CTSZ inhibitor (Cathepsin X‐IN‐1, catalog no. HY‐146985, MCE), or together for 48 h, adding 300 ng/ml LPS simultaneously (for 48 h), and diluting the drug stocks in EMEM with 2% FBS. When 293 APP_Swe cells reached 60% confluence, cells were transfected with (20 nM) siRNAs, and 4 h later were co‐cultured with 30 μM UA, 3 μM CTSZ inhibitor, or both together, diluting the stocks in high‐glucose DMEM with 2% FBS.

2.3. Morris water maze test

The Morris water maze (MWM) test was performed as described previously. ¹¹ The device is a circular pool (140 cm diameter) filled with water and kept at 22°C. The pool was painted with non‐toxic white paint. A transparent platform (12 cm in diameter) was placed 1 cm below the water surface at a fixed position. Mice were trained for 7 days, with four trials per day. Each trial lasted 60 s or until the mouse found the platform. If the mouse did not find the platform within the specified time, the experimenter directed the mouse to the platform. After each trial, the mouse was placed on the platform for 30 s. On the seventh day, after the training phase, the platform was removed for a probe trial, which lasted 60 s. All parameters were recorded by a video tracking system (ANYmaze, version 4.99; Stoelting).

RESEARCH IN CONTEXT

1. Systematic review: Urolithin A (UA) is a gut microbial metabolite of ellagic acid that stimulates mitophagy. We show that UA has a potent anti–Alzheimer's disease (AD) potential in AD mice.

2. Interpretation: In our study, long‐term UA treatment significantly improves learning and memory, olfactory function, and synaptic function of neurons in different AD transgenic mice. It also reduces amyloid beta (Aβ) and tau pathologies. Our data indicate that UA specifically affects the cathepsin Z (Ctsz) protein.

3. Future directions: our study highlights the importance of lysosomal dysfunction in AD etiology, pointing to the high anti‐AD translational potential of UA, a molecule that is clinically safe and bioavailable. Ctsz protein could be a potential therapeutic target for AD treatment.

2.4. Y maze

The Y maze spontaneous alternation performance (SAP) test measures the ability to recognize previously explored environments. ²⁹ The maze consisted of three arms (8 × 30 × 15 cm), with an angle of 120 degrees between them. The number of entries and alterations were recorded by the ANY‐maze video tracking system. Mice were introduced to the center of the Y maze and left to explore the maze for 10 min. Between trials, the arms were cleaned with 70% ethanol solution. SAP is the subsequent entry into a novel arm over the course of three entries; the percentage SAP is calculated by the number of actual alternations/(total arm entries − 2) × 100.

2.5. Object recognition test

The object recognition tests were performed as described previously. ²⁹ The device was an acrylic box of 25 × 25 × 25 cm³. During the training phase, the mice could explore two identical objects for 10 min. During the test phase, 1 h after the training, each mouse was returned to the box, which had been modified to contain a familiar and a novel object. The boxes and objects were cleaned before each test to eliminate scent clues. The automatic video tracking system (ANY‐maze) was used to monitor exploration behavior. Exploration time was calculated as the time each mouse sniffed or pointed its nose or paws at the object. The “recognition index” refers to the time spent exploring the novel object relative to the time spent exploring both objects.

2.6. Olfactory test

The procedure for the buried food test has been described previously. ³⁰ Mice were fasted for 16 h and then placed in a cage containing 3 cm in‐depth bedding with three morsels of food (a small piece of peanut butter flavored cereal; Nestle, Columbia, MD). During the test, each mouse was given a 5‐min period to find the buried food. The latency of the mouse's extraction of the buried food was recorded. The person performing the test and analyzing the data was blinded as to the genotype of the mice.

2.7. Elevated plus maze

Anxiety‐like behavior was measured using the elevated plus maze. ²⁹ The apparatus consists of two closed arms (30 × 5 × 15 cm) with high walls and two open arms (30 × 5 × 2.5 cm) with low walls. Each mouse was placed in the central area of the maze, facing one of the open arms. Time spent in the open arms was measured for 5 min with the EthoVision XT video‐imaging system.

2.8. Open field test

For the open field test, animals were placed in the center of a defined open field region (43 cm × 43 cm) (Med Associates, Georgia, VT, USA) and left without disruption for 20 min. The center zone was defined as a 10.2 cm² spaced equidistant from the peripheral walls. The tracking software (Activity Monitor version 4, Med‐Associates) recorded the exploratory behavior. The apparatus was cleaned with 70% ethanol before testing the next mouse.

2.9. Electrophysiology

Hippocampal slices were prepared as described previously. ²⁹ Briefly, the mice were killed using isoflurane gas, the brains were rapidly removed, and transverse slices were cut at a thickness of 350 μm. Before recording, the slices were allowed to recover for at least 1 h in artificial cerebrospinal fluid (aCSF) at room temperature (RT). aCSF consisted of 120 mM NaCl, 2.5 mM KCl, 1.25 mM NaH₂PO₄, 26 mM NaHCO₃, 1.3 mM MgSO₄, 2.5 mM CaCl₂, and 10 mM glucose (pH 7.4). The osmolarity of the aCSF was adjusted to 290 mOsm using a 5600 Vapor Pressure Osmometer (Wescor, Inc.). Stimuli (30 ms every 20 s) were delivered with a fine bipolar tungsten electrode to activate Schaffer collateral/commissural afferents. LTP was induced with a train of titanic stimulation (100 Hz for 1 s). All recordings were performed at 30–32°C. Data were collected using a MultiClamp 700B amplifier (Molecular Devices). Signals were filtered at 2 kHz, digitized at 10 kHz with a Digidata 1440A Data Acquisition System, and analyzed using pCLAMP 10 software (Molecular Devices).

2.10. Microarray

Gene expression analysis was performed on hippocampal tissue mRNA samples from APP/PS1 and WT littermates. RNA was purified with the Nucleospin RNA isolation kit (catalog no. 740955.250; Macherey‐Nagel), with initial quantitation conducted using a NanoDrop ND‐1000 spectrophotometer (Thermo Fisher Scientific). The quality of the RNA was inspected using a 2100 Bioanalyzer (Agilent Technologies). The microarray was performed by the Gene Expression and Genomics core facility (NIA) and analyzed with JMP, an SAS statistical Suite, combined with R. The initial fluorescent signals were first extracted from the Agilent result files with the background signal testing p values to ensure the validation of the signals. Then, the data were transformed by the log2 function and normalized with quantile normalization. Before further analysis, the data were standardized by z‐transformation to fit for modeling. Then the samples went through the quality evaluation by sample level clustering, sample correlation with scatter plots, and principal component analysis to filter out outliers and ensure the quality of the analysis. The filtered samples were then organized according to their study groups to perform pairwise statistical analysis and variance tests. The z‐test with multiple comparator correction was adopted, along with one‐way analysis of variance (ANOVA). The probes with ANOVA p‐value ≥ 0.05 and background signal‐ranking variance test p‐value ≥ 0.01 were filtered out in the whole set. For each pairwise comparison, we calculated the z‐test p‐value and zratio between pairs. We kept the probes for each pairwise comparison with all of the cutoffs: (1) average signal normalized zcore ≥ 0, (2) z‐test p ≤ 0.05, (3) FDR ≤ 0.30 (we can adjust this according to different needs), (4) |zratio| ≥ 1.5 in the subset probes of our global AVOVA and background filtering. The probes were further averaged on genes for the statistically significant gene list for each pair comparison with their zratios. Over‐representation analysis of the genes that had an FDR of ≤ 0.05 were subjected to KEGG analysis for each pairwise analysis. We required at least three genes in each reported significant term in the data set. All the genes in the respective significant terms were collected.

2.11. Western blotting

Mouse brain tissues were homogenized in 1× RIPA lysis buffer (Cell Signaling, #9806S) containing protease inhibitor cocktail and halt phosphatase inhibitor cocktail (Roche, Indianapolis, IN, USA). Collected samples were sonicated on ice and centrifuged at 10,000 × g for 10 min at 4°C. The protein concentration was determined with a Bradford reagent. Proteins (15 μg) were separated on 4%–15% Bis‐Tris gel (Bio‐Rad Laboratories, #5671085) and transferred to PVDF blotting membranes. Membranes were then blocked for 1 h at RT in TBS‐T (500 mM NaCl, 20 mM Tris, 0.1% Tween 20) supplemented with 5% non‐fat dried milk. Subsequently, membranes were incubated overnight at 4°C with primary antibodies, followed by 1 h at RT with HRP‐conjugated secondary antibodies. Proteins were detected using an enhanced chemiluminescent detection system (EMD Millipore, #WBKLS0500) and ChemiDoc Imaging System (Bio‐Rad Laboratories, #12003153, CA, USA). Quantification was performed using ImageJ. Antibodies used were: Phospho‐Tau (Thr231) (catalog no. MN1040; ThermoFisher); Phospho‐Tau (Thr181) (catalog no. MN1050; ThermoFisher); Phospho‐Tau (Ser202, Thr205) (catalog no. MN1020; ThermoFisher); Tau antibody (catalog no. MN1000; ThermoFisher); Phospho‐NF‐κB p65 (rabbit) (catalog no. 3033; Cell signaling); GFAP antibody (catalog no. Z0334; 1:2000; Dako); Sirt3 antibody (catalog no. 2627; Cell signaling); LAMP1 (catalog no. 21997‐1‐AP; Proteintech); Sirt1 (catalog no. 8469; Cell signaling); PAR (catalog no. 4336‐BPC‐100;Trevigen); BNIP3 (catalog no. 3769; Cell signaling); NIX (catalog no. 12396; Cell signaling); Mfn2 (catalog no. 9482; Cell signaling); ubiquitin (catalog no. 3936; Cell signaling); LC3B (catalog no. NB100‐2220; Novus Biologicals); COXIV (catalog no. 4844; Cell signaling); Pink1 (catalog no. 23274‐1‐AP; Proteintech); Parkin (catalog no. 4211; Cell signaling); NLRP3 (catalog no. 15101; 1:500; Cell signaling); AIM2 (catalog no. 66902‐1‐Ig; Proteintech); STAT3 (catalog no. 10253‐2‐AP; Proteintech); p‐STAT3 (catalog no. 9145; Cell Signaling); Bax (catalog no. 2722; 1:2000; Cell signaling); γH2AX (catalog no. 9718; Cell signaling); Ctsw (catalog no. LS‐C712097‐200; LSBio); Ctsd (catalog no. 21327‐1‐AP; Proteintech); Ctsz (catalog no. 16578‐1‐AP; Proteintech); and β‐Actin (catalog no. A2228; 1:5000; Sigma–Aldrich). Secondary antibodies, including anti‐mouse immunoglobulin G (IgG; 1010‐05) and anti‐rabbit IgG (catalog no. 4010‐05), were obtained from Southern Biotech. Primary antibodies were used at a 1:1000 dilution (unless otherwise stated), with secondary antibodies used at a 1:5000 dilution (unless otherwise stated). All antibodies were validated for use in mouse/human tissues based on previous publications. Detailed antibody validation profiles are available on the websites of the companies from which the antibodies were sourced.

Membranes were then blocked for 1 h at RT in TBST (500 mM NaCl, 20 mM Tris, 0.1% Tween 20) supplemented with 5% non‐fat dried milk and bovine serum albumin (BSA). Subsequently, membranes were incubated overnight at 4°C with primary antibodies, followed by 1 h at RT with HRP‐conjugated secondary antibodies. Proteins were detected using an enhanced chemiluminescent detection system (EMD Millipore, catalog no. WBKLS0500) and ChemiDoc Imaging System (Bio‐Rad Laboratories, catalog no. 12003153, CA, USA).

2.12. Immunofluorescence

The immunofluorescence staining was performed in mouse brain sections, as we reported ¹¹ previously. Pictures were taken using an Axiovert 200 M Zeiss microscope (Zeiss) or a Zeiss 710 LSM confocal microscope (Zeiss). Images were quantified using ImageJ 1.52p. Specific primary antibodies include rabbit anti‐IBA1 (1:500, catalog no. 019‐19741; WAKO) and Ctsz (1:250, catalog no. 16578‐1‐AP; Proteintech).

2.13. Mitochondrial purification

Mitochondrial extraction was adapted and modified from previous studies. ³¹ HMC3 cells with or without Aβ (10 μM) or UA (20 μM) treatment were washed with 1×PBS (phosphate‐buffered saline) and harvested in mitochondria isolation buffer (20 mM HEPES, pH 7.5, 1.5 mM MgCl₂, 1 mM EDTA, 1 mM EGTA, 210 mM sucrose, and 70 mM mannitol). Cell suspensions were subjected to 60 gentle strokes using a Dounce homogenizer. The supernatant obtained was centrifuged at 10,000 g for 15 min at 4°C. The resultant mitochondrial pellet was washed twice and suspended in 200 μL buffer containing 250 mM sucrose, 5 mM magnesium acetate, 40 mM potassium acetate, 10 mM sodium succinate, 1 mM DTT, and 20 mM HEPES/KOH, pH 7.4.

2.14. Enzyme‐linked immunosorbent assay (ELISA) for Aβ and cytokines

Mouse hippocampal and prefrontal cortex (PFC) tissue were homogenized with RIPA buffer containing protease inhibitors, as reported previously. ²⁹ The levels of soluble and insoluble human Aβ₄₀ and Aβ₄₂ in these extracts were quantified using ELISA kits (Thermo Fisher Scientific, catalog no. KHB3442 for Aβ₄₀ and KHB3482 for Aβ₄₂) according to the manufacturer's protocols. The HMC3 cell and 293 APP‐Swedish cell supernatants were analyzed by ELISA according to the manufacturer's instructions. Briefly, cell supernatants were collected in centrifuges; after centrifugation of 5 min at 10,000 g, samples were added into a microplate well and incubated with a fixed amount of target on the solid phase supporter and then added the biotinylated detection antibody specific to the target. Next, Adivin‐HRP was added to each microplate well and incubated. After the TMB solution was added to each well, the enzyme–substrate reaction was terminated by adding a sulfuric acid solution, and the ODs were measured at a wavelength of 450 nm. Human Interlukin 1 beta (IL‐1β) was measured by IL‐1β kit (Thermo Fisher, catalog no. 88‐7261). Human Aβ₄₂ was measured by an Aβ₄₂ kit (ExCell Bio, catalog no. 22A23901).

2.15. RNA extraction and quantitative real‐time PCR

RNA was extracted from mouse brain tissues, as reported previously. ¹⁴ cDNA was synthesized using the PrimeScript RT reagent kit (Takara, catalog no. RR037A), and qPCR analysis was done with a power SYBR Green PCR master mix (Thermo Fisher, catalog no. A46109). The primers used to amplify each transcript were as follows: Ctsz (Forward: 5′‐GGC CAG ACT TGC TAC CAT CC‐3′ and Reverse: 5′‐ACA CCG TTC ACA TTT CTC CAG‐3′), IL‐1β (Forward: 5′‐GCA ACT GTT CCT GAA CTC AAC T‐3′ and Reverse: 5′‐ATC TTT TGG GGT CCG TCA ACT‐3′), IL‐6 (Forward: 5′‐GCC CAG CTA TGA ACT CCT TCT‐3′ and Reverse: 5′‐GAA GGC AGC AGG CAA CAC‐3′), Lamp1 (Forward: 5′‐AGG CCA CTG TGG GAA ACT CAT ACA‐3′ and Reverse: 5′‐TTC CAC AGA CCC AAA CCT GTC ACT‐3′), IL‐10 (Forward: 5′‐CGG GAA GAC AAT AAC TGC ACC C‐3′ and Reverse: 5′‐CGGTTAGCAGTATGTTGTCCAGC‐3′), Trem2 (Forward: 5′‐CTG GAA CCG TCA CCA TCA CTC‐3′ and Reverse: 5′‐CGA AAC TCG ATG ACT CCT CGG‐3′); Tyrobp (Forward: 5′‐GTG ACT TGG TGT TGA CTC TGC TG‐3′ and Reverse: 5′‐GAT AAG GCG ACT CAG TCT CAG C‐3′), Ctsh (Forward: 5′‐ACC GTG AAC GCC ATA GAA AAG‐3′ and Reverse: 5′‐TGA GCA ATT CTG AGG CTC TGA‐3′), Ctsa (Forward: 5′‐CCC TCT TTC CGG CAA TAC TCC‐3′ and Reverse: 5′‐CGG GGC TGT TCT TTG GGT C‐3′), and Gapdh (Forward: 5′‐AGG TCG GTG TGA ACG GAT TTG‐3′ and Reverse: 5′‐GGG GTC GTT GAT GGC AAC A‐3′).

2.16. Gene expression analysis by NanoString Technologies

NanoString analysis was performed on the hippocampus of AD and ADP mice with Veh or UA treatment and their WT littermates. Total RNA was purified with a PureLink RNA Mini Kit (Thermo Fisher Scientific, catalog no. 12183018A) per the manufacturer's protocol. Purified RNA was quantified on a NanoDrop ND‐1000 spectrophotometer and diluted to 20 ng/uL in nuclease‐free water. It was hybridized in CodeSet Master mix carrying hybridization buffer, Reporter Code Set, and Capture Probe Set for 16 to 24 h at 65°C (NanoString Technologies, MAN‐10056‐05) and then applied to the nCounter Prep Station. The Prep Station can process up to 12 samples per run in ≈2.5 to 3 h, depending on the protocol used. We loaded the hybridized RNA onto the nCounter Prep Station for immobilization in the sample cartridge according to the manufacturer's high‐sensitivity protocol (MAN‐C0035). The sample cartridge was subsequently processed for 2.5 h in the nCounter Analysis System. Next, the nCounter Digital Analyzer, a multichannel epifluorescence scanner, collected data by taking images of the immobilized fluorescent reporters in the sample cartridge with a CCD camera through a microscope objective lens. The results were downloaded directly from the digital analyzer in RCC file format. NanoString Advanced analysis (nSolver 4.0) was used for data analysis. Genes with a fold‐change cut‐off of ≥ |1.5| and a p‐value less than 0.05 were considered statistically significant. Pathway terms were considered statistically significant if they had a Gene Set Enrichment score of 1.2.

The NanoString mouse AD panel was analyzed by nSolver Advanced analysis and by ROSALIND (https://rosalind.bio/), with a HyperScale architecture developed by ROSALIND, Inc. (San Diego, CA). Read distribution percentages, violin plots, identity heatmaps, and sample MDS plots were generated as part of the QC step. Normalization, fold changes, and p‐values were calculated using criteria provided by NanoString. ROSALIND follows the nCounter Advanced Analysis protocol of dividing counts within a lane by the geometric mean of the normalizer probes from the same lane. Housekeeping probes for normalization are selected based on the geNorm algorithm implemented in the NormqPCR R library. ³² Fold changes and p values are calculated using the fast method described in the nCounter Advanced Analysis 2.0 User Manual. Venn diagrams were created by Venny 2.1 (https://bioinfogp.cnb.csic.es/tools/venny/). Clustering of genes for heatmaps of differentially expressed genes was done using http://www.heatmapper.ca/expression/. Enrichment was calculated relative to a set of background genes relevant to the experiment; ≥ 1.2 was considered significant.

2.17. Ctsz/Ctsw ELISA assay

Ctsz/Ctsw assays were measured using a mouse ELISA kit (MyBioSource, catalog no. MBS455784 and MBS7218953, San Diego, CA, USA) following the manufacturer's protocol. This assay is based on Ctsz/Ctsw antibody‐Ctsz/Ctsw antigen interactions and an HRP colorimetric detection system to detect Ctsz/Ctsw antigen targets in samples. The hippocampus tissue was homogenized with PBS (pH 7.0 ‐ 7.2) on ice and centrifuged at 10,000 × g for 20 min at 4°C. The supernatants were collected and transferred to the 96‐well clear plate, which was precoated with an antibody specific to Ctsz/Ctsw and had a flat bottom. Brain samples were incubated for 4 h at 37°C and washed twice with 300 μL of 1× wash solution to each well using a multichannel pipette 10 times. Substrate solution to each well was subsequently added and incubated for 15–20 min at 37°C after covering with foil or plate sealer. The enzyme–substrate reaction was terminated by the addition of sulfuric acid (stop solution), and the absorbance value of each well was measured at a wavelength of 450 nm on a microplate ELISA reader (Microplate Manager v5.2.1 software, Bio‐Rad Laboratories).

2.18. Cytokine assay

Mouse eye bleeds were collected in EDTA‐treated tubes. After centrifugation, the supernatant was flash‐frozen. Plasma diluted 1:2 was used to detect cytokines and chemokines using the 31‐plex Cytokine/Chemokine array (Eve Technologies).

2.19. Preparation of Aβ peptides

The peptides were prepared according to the protocols described previously. ¹⁴ Briefly, hexafluoro isopropanol (HFIP)–treated Aβ₄₂ peptides (Chinese peptide) were resuspended in DMSO, and then diluted to a concentration of 100 μM with DMEM/F12 and incubated at 4˚C for 24 h. After centrifugation for 10 min at 14,000 g, the supernatant with soluble Aβ₄₂ was added to culture cells.

2.20. DqBSA assay

HMC3 cells were cultured in EMEM (Merck) containing 10% FBS and 1% GlutaMAX (Thermo Fisher Scientific) with or without 10 μM UA for 6 days. Cells were split, and the following day, dqBSA (Invitrogen, catalog no. D12051) was added in HBSS with or without 10 μM UA or bafilomycin A1 (Tocris Bioscience). dqBSA breakdown was imaged every hour using the Incucyte Zoom live cell analysis system. A basic analysis using the Incucyte Zoom software was conducted to score the total red object integrated intensity (RCU x μm²/image) over time. The linear range was used to determine the degradation rate as was previously done (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6693470/).

2.21. Measurement of lysosomal pH

The LysoSensor Yellow/Blue DND‐160 Kit (Yeasen, #40768ES50) was used according to the manufacturer's instructions. Briefly, HMC3 cells were seeded into a glass bottom cell culture dish and were transfected with (20 nM) siRNAs; 4 h later, they were co‐cultured with 10 μM Aβ₄₂, 30 μM UA, and 3 μM CTSZ Inhibitor or together for 48 h, diluting the drug stocks in EMEM with 2% FBS, and loaded with 3 μM LysoSensor Yellow/Blue DND‐160 of MEM for 30–40 min at 37˚C. Then the cells were washed three times with PBS buffer and immediately observed by fluorescence microscopy (OLYMPUS).

2.22. Randomization and blinding

Animal/samples (mice) were assigned randomly to the various experimental groups, and mice were randomly selected for the behavioral experiments. In data collection and analysis (for example, mouse behavioral studies and mouse imaging data analysis), the performer(s) were (were) blinded to the experimental design.

2.23. Statistical analysis

Prism 9.0 (GraphPad Software) was used for the statistical analysis. Data shown are the mean ± standard error of the mean (SEM) with p ≤ 0.05 considered statistically significant. Two‐tailed unpaired t‐tests were used for comparisons between the two groups. Group differences were analyzed with one‐way ANOVA followed by Tukey's multiple comparisons test or two‐way ANOVA followed by Tukey's multiple comparisons test for multiple groups. No statistical methods were used to predetermine sample sizes, but our sample sizes (mouse experiments) were similar to those reported in previous publications.

2.24. Data availability

The microarray GEO accession numbers for the data reported in this paper are GSE212972, GSE214406, and GSE214416. All data are available from the corresponding author upon reasonable request.

3. RESULTS

3.1. Early and long‐term UA treatment improves learning and memory in APP/PS1 mice

APP/PS1 mice are used widely in AD research, and they develop Aβ pathology and behavioral deficits by 6 months. ²¹ , ³³ , ³⁴ Thus we sought to initiate UA treatment before significant AD phenotypes occur. Notably, mitophagy defects may occur before the main pathologies of AD. Therefore, early intervention testing with mitophagy stimulation is essential. Two‐month‐old APP/PS1 mice were exposed to UA for 5 months, and behavioral tests were performed. Mouse brains were collected (Figure 1A). In the MWM experiment, we found that APP/PS1 mice took more time to reach the hidden platform than WT mice, indicating that their learning ability had decreased (Figure 1B). Compared with APP/PS1 mice treated with vehicle, APP/PS1 mice treated with UA took significantly less time to reach the platform, like that of WT (Figure 1B). This result shows that UA administered for 5 months significantly improved the learning ability of APP/PS1 mice. During the experiment, the swimming speed of mice in each group was similar (Figure 1C), demonstrating that UA improved learning ability independent of motor function. In addition, we used another cognitive function test, Y‐maze, to evaluate the effect of long‐term UA treatment on memory. Five‐month UA treatment significantly improved the memory ability of APP/PS1 mice (Figure 1D).

FIGURE 1.

UA improved learning and memory ability in APP/PS1, 3xTgAD (AD), and 3xTgAD/Polβ^+/− (ADP) mice. (A) Experimental design for UA‐treated APP/PS1 mice. (B) The learning phase of Morris water maze. The latency to find the hidden platform of WT, APP/PS1, and UA‐treated APP/PS1 mice. **P < 0.01, APP/PS1 mice compared to WT mice. #P < 0.01, UA‐treated APP/PS1 mice compared to vehicle‐treated APP/PS1 mice. For learning curve comparison: P = 0.002, APP/PS1 compared to WT; P = 0.0021, APP/PS1+UA compared to APP/PS1, by Pearson's correlation; n = 9–10 male mice. (C) The swimming speed in Morris water maze, n = 9–10 male mice. (D) Y‐maze test of 5‐month UA‐treated APP/PS1, vehicle‐treated APP/PS1 mice, and WT mice; n = 9–10 male mice. (E) Y‐maze test of 5‐month UA‐treated APP/PS1, vehicle‐treated APP/PS1 mice, and WT mice, after stop administration of UA for 1 month; n = 4–5 male mice. (F) Experimental design for UA‐treated AD and ADP mice. (G) Y‐maze test of 5‐month UA‐treated or vehicle‐treated AD, ADP, and vehicle‐treated WT mice; n = 6–9 mice. (H, I) Object recognition test of 5‐month UA‐treated or vehicle‐treated AD, ADP, and vehicle‐treated WT mice; n = 6–9 mice. (J) The latency to find cookies in the buried food test of the UA‐treated or vehicle‐treated AD, ADP, and vehicle‐treated WT mice; n = 7–9 mice. (K) Elevated plus maze of the UA‐treated or vehicle‐treated AD, ADP, and vehicle‐treated WT mice; n = 7–9 mice. (L) Open field test of the UA‐treated or vehicle‐treated AD, ADP, and vehicle‐treated WT mice. (M, N) Results of analyses of LTP measured at the Schaffer collateral synapses of the UA‐treated or vehicle‐treated AD, ADP and vehicle‐treated WT mouse brains. Values are the mean and SEM of determinations made on 5–7 hippocampal slices from at least five different mice. EPSP, excitatory postsynaptic potential. Data were analyzed by two‐way ANOVA with Tukey's multiple comparisons test and Pearson's correlation (B) or one‐way ANOVA with Tukey's multiple comparisons test (C–E, G–L, N). Data were shown as mean ± SEM. The APP/PS1 mice are all male. For G–I, WT (7 M and 2 F), AD (3 M and 4 F), AD + UA (3 M and 3 F), ADP (5 M and 3 F), ADP + UA (3 M and 5 F); for J–L, WT (7 M and 2 F), AD (3 M and 4 F), AD + UA (3 M and 4 F), ADP (5 M and 3 F), ADP + UA (4 M and 5 F). F, female, M, male.

We wondered if UA treatment would have a lasting impact on disease endpoints, so we measured Y‐maze 1 month after discontinuing the 5‐month administration of UA. After 1 month without UA, APP/PS1 mice treated with UA still showed improved memory ability compared to APP/PS1 mice treated with vehicle (Figure 1E). Motor function and anxiety were not changed in the APP/PS1 mice compared to WT mice, or after UA treatment (Figure S1A–D). These results show that in 2‐month‐old APP/PS1 mice given 5 months of UA, learning and memory function improved significantly, and this effect persisted 1 month after UA was discontinued.

3.2. Long‐term UA treatment improved learning and memory in 3xTgAD and 3xTgAD/Polβ^+/− mice

To verify the therapeutic effects of early and long‐term UA treatment on AD, we used two other AD models: the widely used AD and the DNA repair deficient ADP mice, which we developed. ²⁸ The 3xTgAD mice are APP and tau transgenic mice, and we crossed them with DNA repair‐deficient Polβ^+/− mice ²⁸ and reported that they had many more human AD features than the 3xTgAD mice. ²⁸ Decreased DNA repair activity has been found in AD patients, and we reported a base excision DNA repair defect at the level of DNA Polβ in human early postmortem AD brains. ³⁵ In our AD and ADP mice, we typically find Aβ and tau pathology starting at around 12 months. ²⁸ For the AD and ADP mouse models, UA was administered orally from 12 to 17 months (duration 5 months, Figure 1F). UA significantly improved the memory of AD and ADP mice in the Y‐maze test (Figure 1G). In the object recognition test, the cognitive ability of AD and ADP mice to recognize two identical objects was like that of WT mice (Figure 1H), but the ability to recognize novel objects had significantly decreased (Figure 1I). After UA treatment, the novel object recognition ability of AD and ADP mice had improved significantly (Figure 1I), showing that UA improved the cognitive ability of AD and ADP mice.

Sensory perception is altered in AD patients. AD patients commonly have olfactory impairment, ³⁶ and we have reported previously that our ADP mice have olfactory dysfunction. ³⁷ , ³⁸ Thus we tested the mice using the buried food smelling test. The latency of ADP mice to find the buried cookies was significantly longer than that of WT, and long‐term treatment with UA improved the olfactory function in these mice, whereas the AD mice did not have a significant smelling impairment (Figure 1J).

To assess UA's effects on anxiety, we first used the elevated plus maze test. AD and ADP mice had no anxiety‐related behavior in this experiment and were not significantly different from WT (Figure 1K). However, in the open field test, the walking distance of the AD and ADP groups was reduced relative to WT mice, and the walking distance of the UA‐treated AD group was shorter than that of the AD group. However, this effect was not observed in the ADP mice group (Figure 1L and Figure S1E,F). To further compare the effects of UA on the behavior of AD model mice by sexes, we divided female and male mice for statistical analysis. The results showed that female and male mice had similar trends (Figure S1E,F, right panels; Figure S2).

The decline of cognitive function in AD patients is related to the loss of synaptic function. Therefore, we studied long‐term potentiation in the mice using electrophysiological experiments. The results showed that UA significantly improved the long‐term potentiation of AD and ADP mice (Figure 1M,N). Collectively, these results indicate that long‐term administration of UA improves the cognitive, olfactory, and synaptic functions of neurons in AD and ADP mice.

3.3. UA decreased Aβ accumulation and tau phosphorylation in AD mice

To investigate the effects of UA treatment on Aβ and phosphorylated p‐tau, we used APP/PS1, AD, and ADP mice. In APP/PS1 mice, we performed ELISAs to examine whether UA treatment decreased the levels of Aβ in the hippocampus and prefrontal cortex (PFC), known as two of the earliest and most affected brain areas in AD. ³⁹ There were no significant changes in insoluble and soluble of Aβ₄₂ and Aβ₄₀ in the hippocampus of APP/PS1 mice after long‐term UA treatment (Figure 2A–D). However, in the PFC, both insoluble and soluble Aβ₄₂ were significantly decreased after long‐term UA treatment (Figure 2A,C). The levels of soluble Aβ₄₀ were also reduced in the PFC of APP/PS1 mice (Figure 2D), whereas insoluble Aβ₄₀ showed no changes (Figure 2B).

FIGURE 2.

UA decreased Aβ accumulation and tau phosphorylation in AD mice. (A–D) The effects of UA treatment on Aβ productions in both hippocampus and PFC. Insoluble Aβ₄₂ (A), insoluble Aβ₄₀ (B), soluble Aβ₄₂ (C), and soluble Aβ₄₀ (D) levels in all groups. Data were shown as mean ± SEM (n  =  3 male mice in the AD group; n  =  5 male mice in the AD+UA‐washout group; n  =  4 male mice in all the other groups; *P  <  0.05; two‐way ANOVA with Tukey's multiple comparisons test). (E, F) The representative images and quantification of immunostaining of IBA1 in the cortex and hippocampus in WT, vehicle‐, or UA‐treated APP/PS1 mice; n = 3 male mice per group. Scale bars, 100 and 400 μm, as indicated. (G) Western blot data showing the effects of UA treatment on the expression level of proteins involved in Aβ accumulation and tau phosphorylation in the hippocampus of WT, 3xAD, and 3xAD/Polβ^+/− mice; n = 3 mice per group. (H) Quantification of protein levels for p‐tau (Thr181), p‐tau (Thr231), p‐tau (Ser202, Thr205), and total tau. (I, J) Quantification of protein levels for GFAP and Sirt3. For H‐J, n = 3 mice per group. All samples were normalized to respective loading controls β‐actin. Data were analyzed by one‐way ANOVA with Tukey's multiple comparisons test (I, J) or two‐way ANOVA with Tukey's multiple comparisons test (A–D, F, H). Data were shown as mean ± SEM. The APP/PS1 mice are all male. For AD, AD+UA: 2 M and 1 F; For WT, ADP and ADP+UA, 1 M and 2 F. M, male; F, female.

One month after discontinuing UA administration, no significant changes were observed for Aβ₄₂ and Aβ₄₀ in the hippocampus of APP/PS1 mice (Figure 2A,C). We also did not detect decreased levels of Aβ₄₀ and Aβ₄₂ in the PFC in the group without UA administration for 1 month (Stop administration, SA), suggesting that the beneficial effects observed in the PFC vanished in APP/PS1 mice 1 month after stopping UA administration (Figure 2A–D, SA group). The results show that UA treatment reduced Aβ accumulation in the PFC but not in the hippocampus of APP/PS1 mice and that the effects disappeared if UA treatment was suspended for 1 month.

Abnormally activated neuroinflammation is an important pathological feature of AD. We detected activated microglia cells in the brains of AD mice. The number of IBA1‐positive cells increased significantly in the cortex and hippocampus of APP/PS1 mice, and UA reduced these abnormally activated microglia cells dramatically in the cortex and hippocampus of APP/PS1 mice (Figure 2E,F).

Next, we investigated the effects of UA treatment on tau phosphorylation in ADP brains. Previous studies have shown that phosphorylation sites of tau, including Thr181, Thr231, and Ser202/Thr205, were associated with disease progression. ⁴⁰ Notably, UA treatment significantly decreased total tau levels at Thr181 and Thr231 in both AD and ADP mice (Figure 2G,H). UA treatment also decreased p‐tau levels in ADP but not in AD brains (Figure 2G,H). Thus our results showed that UA treatment suppressed the accumulation of total tau and p‐tau in brains.

Glial fibrillary acidic protein (GFAP), a marker for astrocytes, is associated with AD pathology as an early marker of brain Aβ status. ⁴¹ We found that GFAP was remarkably increased in AD brains and normalized after UA treatment (Figure 2G,I). However, there were no differences in the expression of GFAP between ADP and WT mice. It was reported that the neurotoxic Aβ peptides induced NF‐κB activation in AD brains. ⁴² Because inflammation is a major cause of Aβ and p‐tau, ⁴³ we explored whether UA treatment attenuated the inflammatory reactions in AD mice. Of interest, long‐term UA treatment tended to decrease the expression level of activated NF‐κB, as determined by increased p65 in the hippocampus of AD mice compared with WT mice (Figure 2G and Figure S3A). However, there was no difference between WT and AD/ADP mice without UA treatment, suggesting that the downregulated NF‐kB responded specifically to UA treatment (Figure 2G and Figure S3A). In addition, sirtuin 3 (SIRT3) is a mitochondrial deacetylase. Its dysfunction is strongly associated with the pathogenesis of AD. ⁴⁴ A previous study revealed that Aβ increased total tau levels through regulating SIRT3 in AD adults. ⁴⁵ It is notable that our results indicated that Sirt3 tended to decrease in the ADP mice but not in the AD mice (Figure 2G,J). Moreover, the expression level of Sirt3 significantly increased after long‐term UA treatment in ADP mice (Figure 2J). Taken together, we speculate that UA might downregulate Aβ‐ and p‐tau‐induced inflammation and DNA damage in AD brains.

3.4. UA restored lysosomal functions in part by regulating cathepsin Z

To further explore the underlying mechanisms related to UA treatment of AD, we performed gene expression analysis by microarray using hippocampal tissues from WT, untreated, and treated APP/PS1 (AD) mice. KEGG terms showed that the most significantly altered term in UA‐treated AD mice was related to lysosomes (Figure 3A). The significantly differentially expressed genes are shown in Figure 3B. Cathepsin A (Ctsa), cathepsin H (Ctsh), cathepsin L (Ctsl), cathepsin S (Ctss), and cathepsin Z (Ctsz) were all significantly upregulated in AD brains, and UA treatment decreased their expression (Figure 3B). Cathepsins are vital lysosomal compartment enzymes responsible for intracellular protein degradation, energy metabolism, and immune responses. ⁴⁶ Dysregulation of cathepsin proteases may contribute to the development of AD and Aβ production. ⁴⁷ Consistent with these results, our data showed a specific increase in cathepsin gene expression levels in AD brains, which might contribute to the progression of AD (Figure 3B). The mRNA levels of the cathepsins (Ctsa, Ctsh, Ctss, Ctsz) were examined by quantitative real‐time polymerase chain reaction (qPCR) in all brain samples (Figure 3C, Figure S3B–D).

FIGURE 3.

UA restored lysosomal functions by regulating Ctsz in APP/PS1 mice. (A) KEGG term analysis of gene expression microarray from the hippocampus of APP/PS1 mice. The most changed terms in AD compared to WT hippocampus were shown. (B) The FDR‐selected genes (cutoff 0.05) from each term in (A) from the hippocampal microarray of APP/PS1 compared to WT and APP/PS1 compared to APP/PS1 with UA treatment are shown. (C) qPCR analysis for the relative gene expression of Ctsz. (D) Measurement of Ctsz levels by ELISA assay in the hippocampus of APP/PS1 or WT mice with and without UA treatment. (E, F) Western blots of the Ctsz protein expression levels in the hippocampus of APP/PS1 with or without UA treatment. (G) qPCR analysis for the relative gene expression of Lamp1 in the hippocampus. For qPCR, n = 3–4 mice. (H) The representative images of immunostaining of Ctsz in the hippocampus and cortex of WT, vehicle‐, or UA‐treated APP/PS1 mice; n = 3 mice per group. Scale bars, 400 and 100 μm, as indicated. (I) The quantification of Figure 3H. (J) DqBSA degradation assay in HMC3 cells with UA treatment or Baf treatment. (K–M) Representative western blots showing LAMP1 and CTSZ and quantitation of protein levels in UA or Aβ or UA + Aβ‐treated HMC3 cells, n = 9 independent repeats. (N) ELISA for Aβ₄₂ in UA or CTSZ inhibitor or si‐CTSZ treated 293‐APPsw cell supernatant. (O, P) The measurement of lysosomal pH by using the LysoSensor Yellow/Blue DND‐160 Kit, HMC3 cells were treated with siRNAs, 10 μM Aβ₄₂, 30 μM UA, and 3 μM CTSZ Inhibitor or together for 48 h. Scale bars, 200 μm. Data were shown as mean ± SEM. Data were analyzed by one‐way ANOVA with Tukey's multiple comparisons test (C, D, F, G, L–N, P) or two‐way ANOVA with Tukey's multiple comparisons test (I, J). All APP/PS1 strain mice used are males.

Among them, we found that Ctsz was significantly elevated in the hippocampus of AD brains, and UA normalized its expression to control levels (Figure 3C). CTSZ is a member of the lysosomal cysteine cathepsin protease family, which comprises 11 members in humans. ⁴⁶ AD patients have been reported to have increased CTSZ expression. ⁴⁸ CTSZ immunoreactivity in AD patient brains has also been reported, however, in less detail and using a different primary antibody. ⁴⁹ In the context of neuroinflammation, Ctsz has been implicated in the development of inflammation, Interleukin 1 beta (IL‐1β) production, and NLR Family Pyrin Domain Containing 3 (NLRP3)–inflammasome activation. ⁵⁰ , ⁵¹ We measured Ctsz levels by ELISA to further evaluate the effects of UA in AD mouse brains. The Ctsz level was significantly higher in the hippocampus of APP/PS1 mice than in WT mice and decreased notably after UA treatment (Figure 3D).

The mRNA expression levels of Ctsa and Ctsh were higher in the hippocampus of APP/PS1 mouse brains than in controls, but UA did not affect them significantly (Figure S3B,C). Ctss was not changed in the AD and UA‐treated groups compared with WT (Figure S3D). The protein levels of Ctsz were increased in the AD mouse hippocampus and decreased after UA treatment (Figure 3E,F). We additionally used immunostaining and found that Ctsz was higher in the cortex and hippocampus of the APP/PS1 mice and decreased after UA treatment (Figure 3H,I). The lysosomal marker, lysosomal‐associated membrane protein 1 (Lamp1), was more highly expressed at the mRNA level in the hippocampus of AD mice and decreased significantly after UA treatment, suggesting that UA may contribute to downregulating lysosomal degradation in AD (Figure 3G).

To examine lysosomal degradation activity after UA treatment, we performed the Dye Quenched–Bovine Serum Albumin (DqBSA) assay and the lysosensor assay. In the DqBSA degradation assay, UA significantly increased DqBSA degradation in the HMC3 cells, and bafilomycin (Baf) was used as a negative control (Figure 3J), suggesting that UA increased lysosomal function.

Microglia are major immune cells of the central nervous system (CNS), and HMC3 human microglia were used for further study. HMC3 human microglia were treated with Aβ₄₂ (10 μM) with or without UA for 48 h, and the protein levels of CTSZ and LAMP1 decreased when UA was added with Aβ₄₂ (Figure 3K–M). In addition, we used HEK293‐APPsw cells stably expressing the APPsw mutant to investigate the expression of Aβ after different treatments. ELISA showed that UA decreased Aβ₄₂ expression in the supernatant of HEK293‐APPswe cells (Figure 3N). Using the lysosensor kit to detect lysosomal pH, we found increased yellow DND160 signal in the cells treated with UA, suggesting that UA promotes acidification of lysosomes and thereby may improve lysosomal function (Figure 3O). Note, Aβ treatment reduced the cellular lysosome activity, and UA restored it. Of interest, the UA‐induced lysosomal activity was blocked after CTSZ knockdown, maybe because CTSZ knockdown activates lysosome acidification by itself, thereby rendering it more difficult to see the stimulation of UA, or implicating that UA can improve lysosome function and this function may be partially mediated by CTSZ (Figure 3O,P). CTSZ knockdown efficiency is shown in Figure S3E,F.

3.5. UA decreased inflammation in AD mice

CTSZ is associated with neurodegenerative diseases like AD and Huntington's disease (HD) ⁵² , ⁵³ and is related to inflammation and IL‐1β production. ⁵¹ We next examined the mRNA expression level of IL‐1β and Interleukin 6 (IL‐6) in the hippocampus to assess the effects of UA on inflammation. IL ‐1β was enhanced in AD mice, and UA notably decreased its level (Figure 4A). However, there was no difference in IL‐6 expression among all groups (Figure 4B), suggesting that UA weakened inflammation mainly through the IL‐1β‐related signaling pathway. IL‐10 is an anti‐inflammatory cytokine, which was decreased in AD mice and restored after UA treatment (Figure 4C). Cytokine levels were also measured from cortex lysates using the Cytokine assay. We found that several proinflammatory cytokines were increased in AD mice and decreased after UA treatment, including Interleukin 1 alpha (IL‐1α), monocyte chemoattractant protein‐1 (MCP‐1), macrophage inflammatory protein‐1 alpha (MIP‐1α), tumor necrosis factor (TNFα), Interleukin 2 (IL‐2), and keratinocyte‐derived cytokine (KC) (Figure 4D–I). The IL‐6 level did not show significant differences between the groups (Figure S3G). Some other cytokines were changed, and most of them showed the same trends: increased in AD mice cortex lysates and decreased in UA‐treated mice (Figure S3H–S). To further investigate the mechanism by which UA enhances lysosome function and the correlation with CTSZ, HMC3 cells primed with LPS were treated with CTSZ‐siRNA or CTSZ inhibitors, with or without Aβ₄₂ and UA. In the Aβ₄₂‐treated condition, UA significantly reduced IL‐1β expression in the HMC3 cell supernatant but did not reduce IL‐1β in the CTSZ‐knockdown. Consistent with the CTSZ knockdown results, CTSZ inhibitors also reduced IL‐1β expression, and UA did not further reduce IL‐1β in HMC3 cells treated with CTSZ inhibitors. These results suggest that UA can reduce IL‐1β, and inhibition of CTSZ can also reduce IL‐1β, and the way UA reduces IL‐1β may be partly dependent on CTSZ (Figure 4J). However, IL‐1β is only one inflammatory factor, so it does not fully represent neuroinflammation. The in‐depth mechanism remains to be further explored in the future.

FIGURE 4.

UA decreased proinflammatory factors in AD mice. (A–C) qPCR analysis for the relative gene expression of IL‐1β (A), IL‐6 (B), and IL‐10 (C). (D–I) The cytokine assay of cytokine levels in mice cortex lysates, including IL‐1α, MCP‐1, MIP‐1α, TNFα, IL‐2, and KC, n = 5 in the WT group; n = 3 in the APP/PS1 group; n = 4 in the APP/PS1 + UA group. (J) ELISA of IL‐1β in Aβ₄₂ or UA or CTSZ‐knockdown or CTSZ‐inhibitor treated HMC3 cells; n = 3 biological duplicates. (K, L) qPCR analysis for the relative gene expression of Trem2 (K) and Tyrobp (L). For qPCR, n = 3–4 mice. Data were shown as mean ± SEM. Data were analyzed by one‐way ANOVA with Tukey's multiple comparisons test (A–L). For APP/PS1 strain mice, all used are males.

The triggering receptor expressed on myeloid cell 2 (TREM2) binds to the tyrosine motif binding protein (TYROBP; also known as DAP12, DNAX‐binding protein‐12), a critical modulator in microglial biology and correlates with a high risk of AD development. ⁵⁴ , ⁵⁵ Our RNA sequencing data by NanoString showed that the expression levels of Trem2 and Tyrobp markedly increased in APP/PS1 mice and decreased dramatically after UA treatment in APP/PS1 brains (Figure 3B), and this was verified by the mRNA expression results from the hippocampus tissues (Figure 4K,L). CD68 is a lysosomal protein expressed at high levels by macrophages and activated microglia. We found that CD68 was higher in the hippocampus of APP/PS1 mice and unchanged in UA‐treated APP/PS1 mice (Figure S4A). To investigate the effects of UA on neuroinflammation, we tested some inflammation‐related proteins in mouse hippocampus and found that UA decreased Nuclear Factor kappa B (NF‐κB) p105/p50, Absent in Melanoma 2 (AIM2), and phosphorylated signal transducer and activator of transcription 3 (pSTAT3)/STAT3 (Figure S4B–F). The mTOR pathway has been implied in the regulation of autophagy and mitophagy. ⁵⁶ We found that the mTOR expression level was decreased in APP/PS1 mouse brains and increased after UA treatment (Figure S4G,H). UA also trended to decrease CD68 protein in APP/PS1 mice brains (Figure S4G,I). Accordingly, these data suggest that UA administration reduces neuroinflammation in AD mice.

3.6. UA induced sirtuin expression, mitophagy, and decreased DNA damage

The sirtuin family of proteins impacts both aging and mitophagy. ⁵⁷ Thus, we assessed the expression of Sirt1 and Sirt3 in the AD mouse cortex and found that UA increased the expression of both proteins in APP/PS1 mice (Figure 5A–C). In our previous study, we found that UA plays an important role in AD therapy as a mitophagy inducer. ¹¹ Here, we also verified the effect of long‐term administration of UA on mitophagy‐related indicators in APP/PS1 mice. UA increased the expression of Parkin and BCL2 interacting protein 3 (BNIP3) (Figure 5D and Figure S5A,E), which were decreased in APP/PS1 mice, and tended to increase the expression of other mitophagy‐related proteins, such as Nix and Mitofusin 2 (Mfn2) (Figure 5D and Figure S5B,C). Long‐term administration of UA was also found to dramatically reduce DNA damage responses, as assessed by phosphorylated H2AX, γ‐H2AX (Figure 5D,F). We also investigated PARylation level and p‐Parkin in UA‐treated APP/PS1 mouse cortex samples and found that UA decreased PARylation and p‐Parkin in APP/PS1 mouse cortex samples, supporting that UA decreases DNA damage responses (Figure S5D–F).

FIGURE 5.

UA modulates both general immune and neuronal processes and AD‐specific pathophysiological pathways. (A–C) Representative western blots showing Sirt1 and Sirt3 and quantitation of protein levels in the cortex of WT, APP/PS1, and APP/PS1 with UA mice; n = 3 male mice per group. (D–F) Representative western blots showing PAR, Parkin, BNIP3, NIX, Mfn2, γH2AX, and quantitation of protein levels in the cortex of WT, APP/PS1, and APP/PS1 with UA mice; n = 3 male mice per group. (G) The representative western blots showing ubiquitin, Parkin, PINK1, LC3B, and COXIV in the HMC3 cell of NC, UA, Aβ and Aβ + UA treatment; n = 3 independent repeats. (H) The heatmap in the hippocampus shows gene set enrichment analysis of significant terms from the various AD clusters defined by the NanoString mouse AD panel. Cluster designations A–E are shown to the left and defined below. Cluster A represents ECM organization. Cluster B represents the immune system. Cluster C represents the neuronal system. Cluster D, cell cycle, NMD. Cluster E represents organelle biogenesis, cell stress response. (I) Heatmap in the in hippocampus showing gene set enrichment analysis of significant pathology terms defined by the NanoString mouse AD panel. (J) Venn diagram of AD and ADP mice showing significant genes (p‐value ≤ 0.05) and their relative overlap between the treatment groups. (K) Set of significantly changed genes (fold‐change ≥ |1.2| and p‐value ≤ 0.05) changed similarly by UA in both AD and ADP. Data were shown as mean ± SEM. Data were analyzed by one‐way ANOVA with Tukey's multiple comparisons test (B, C, E, F).

Alterations in mitochondria and lysosomes are often both present in neurodegenerative diseases, suggesting a close relationship between mitochondria and lysosomes. ⁵⁸ To investigate the effect of UA on mitophagy, we examined the markers of mitophagy in mitochondrial extracts. We used UA, Aβ, and Aβ + UA in HMC3 cells. Mitochondria were extracted, and proteins related to mitochondrial autophagy were assessed. Mitochondrial Parkin was significantly increased, and ubiquitin, PTEN‐induced kinase (PINK1), and microtubule‐associated proein 1A/1B‐light chain 3 beta (LC3B) tended to increase after Aβ treatment, and UA restored the expression of these mitophagy‐related proteins to similar levels as in the control group (Figure 5G and Figure S6A–D). The results indicate that acute Aβ stimulation leads to abnormal mitophagy‐related protein expression in extracts, and UA restores these proteins to normal levels.

3.7. Targeting multiple mechanisms of pathology, UA modulates immune responses and AD‐specific pathophysiological pathways

To assess gene expression changes after UA, we employed the NanoString Alzheimer's Disease panel (Figure 5H). The AD panel was designed to correlate key human disease processes and pathways with mRNA from mouse brains. It consists of genes associated with the primary molecular characteristics of AD. ⁵⁹ The genes included 30 AD‐associated brain region‐specific gene co‐expression modules defined by the Accelerating Medicines Partnership–Alzheimer's Disease (AMP‐AD) consortium and 23 neurodegenerative pathways and processes. ⁵⁹ , ⁶⁰ The co‐expression modules are further partitioned into five distinct consensus clusters that share minimal gene overlap (clusters A–E). Each cluster is associated with various concepts such as A, extracellular matrix organization; B, immune system; C, neuronal system; D, cell cycle and nonsense mediated decay; E, organelle biogenesis and cell cycle response. Consistent with the previous article, we set our Gene Set Enrichment Analysis threshold for significance at 1.2. ⁶⁰

When compared to WT vehicle‐treated mice, both 3xTgAD (AD) and 3xTgAD/Polβ^+/− (ADP) models showed upregulation of clusters A–D and downregulation of cluster C, with our ADP model repeatedly showing greater deviation from WT mice (Figure 5H). Our mice data are consistent with other AD brain transcriptomes and various mouse models, wherein cluster B genes are upregulated and cluster C genes are predominately downregulated. ⁶⁰ The response to UA was consistent in both AD and ADP mice, as UA induced downregulation of clusters A–D and upregulation of cluster C. Notably, many cluster B and D terms were normalized by UA treatment in AD mice, whereas they were only mildly downregulated in ADP mice. Aging is the strongest risk factor for AD, and cluster B terms are associated with immune function and are activated by age and in several other neurodegenerative and neuropsychiatric disorders. ⁶⁰ In contrast, oligodendroglial‐enriched cluster D terms, Fpblue and TCXyellow, which are activated in select AD models but rarely in other neurodegenerative disease models, ⁶⁰ were consistently downregulated by UA, suggesting that UA may be targeting multiple mechanisms of pathology and modulate general immune, neuronal processes, and AD‐specific pathophysiological pathways.

Concerning the AD neurodegenerative pathways and processes, most of the significant terms were upregulated in both AD and ADP models relative to WT mice, and again, our ADP model showed greater deviation from WT than AD (Figure 5I). UA treatment induced downregulation of myelination, cytokines, angiogenesis, tissue integrity, and growth factor signaling in both genotypes. Vesicle trafficking, trophic factors, and transcription and splicing were downregulated terms in the AD models relative to WT mice. UA caused the upregulation of trophic factors and vesicle trafficking terms. Of interest, UA downregulated oxidative stress in AD mice, but there was no change in ADP mice. Based on gene expression findings and behavioral studies, ²⁸ , ²⁹ , ³⁸ ADP mice possess more advanced AD pathology, and perhaps UA is less efficacious in advanced disease states.

Because there are only 760 genes on the NanoString AD panel, we collected all the genes from each pairwise comparison with a p‐value of ≤ 0.05 and compared them in a three‐way Venn diagram (Figure 5J). There were 155 genes within the AD strain and 245 genes within the ADP strain comparisons. A majority of the genes in each pairwise comparison were unique to that comparison. To identify genes with a common UA response in AD and ADP, we combined the gene lists from the two strains and then selected those genes that were significantly changed (defining a fold change ≥ |1.2| in any comparison as significant). This limited the gene list to 46, which is insufficient to do meaningful pathway analysis. Thus, we then sought to identify genes that were similarly significantly changed in the AD_WT or ADP_WT comparisons that were normalized by UA treatment (Figure 5K). UA downregulated plasma membrane calcium transporting ATPase 1, Atp2b1; AF4/FMR2 family member 1, Aff1; ras homolog gene family member U, Rhou; dysbindin domain containing 2, Dbndd2; and zinc finger E‐box binding homeobox 2, Zeb2. UA also upregulated carbonic anhydrase 10, Car10; nerve growth factor inducible, Vgf; ligand of numb‐protein X 1, Lnx1; and cAMP‐dependent protein kinase type I regulatory subunit Beta, Prkar1b. Among these nine genes, dysregulation of calcium in neurodegenerative disorders is common, and Atp2b1 removes calcium ions from eukaryotic cells. Furthermore, ATP2B1 was identified previously as a hub gene in a transcriptomics study of human AD postmortem prefrontal cortical tissue. ⁶¹ VGF is also notable because a recent study identified VGF as a “high”‐confidence master regulator of AD‐associated networks. ⁶² It is downregulated in human AD patients ⁶³ and its expression correlates with disease progression. Similarly, it was downregulated in our mice and notably upregulated by UA. Overexpression in the 5xFAD mouse model partially rescued AD pathologies, including memory impairment. ⁶² How this short list of genes contributes to the altered AD pathology is beyond the scope of the study here but warrants further study.

3.8. UA decreases neuroinflammation, mitochondrial markers, and DNA damage

UA is the most active, effective gut metabolite shown to stimulate mitophagy and acts as a potent anti‐inflammatory and antioxidant agent. ⁶⁴ To further investigate the beneficial effects of UA, the expression levels of important markers of mitochondrial stress, neuroinflammation, and DNA damage were explored in the hippocampus from Veh‐ and UA‐treated 3xTgAD (AD) and 3xTgAD/Polβ^+/− (ADP) mice (Figure 6A–G). In the absence of stress, PINK1 is constitutively imported and undergoes cleavage by the mitochondrial processing peptidase and presenilin‐associated rhomboid‐like protease in the inner mitochondrial membrane. ⁶⁵ , ⁶⁶ Upon mitochondrial damage, PINK1 is no longer imported; full‐length PINK1 accumulates on the outer mitochondrial membrane and leads to autophosphorylation, which promotes kinase activation and facilitates binding to substrates Parkin and ubiquitin. ⁶⁷ , ⁶⁸ As shown in Figure 6A, our results showed that the accumulation of full‐length Pink1 and Parkin was markedly increased compared with the WT group, accompanied by a significant increase in Bax protein expression in the hippocampus of AD and ADP mice. This protein expression were notably normalized by UA treatment, except for the Parkin level in AD mice (Figure 6B–D). Next, we examined neuroinflammation‐related protein expression to evaluate the effects of UA on neuroinflammation. NLRP3 and AIM2 inflammasome complexes are critical components of the innate immune system that mediate caspase‐1 activation and induce proinflammatory cytokine production. ⁶⁹ After activation of the NLRP3 inflammasome, the secretion of inflammatory cytokines such as IL‐1β and tumor necrosis factor alpha (TNFα) leads to a strong inflammatory response, impairs nerve cell function, and consequently leads to cognitive dysfunction. ⁶⁹ AIM2 is also a critical inflammasome sensor that recognizes cytosolic double‐stranded DNA (dsDNA) via its HIN200 domain ⁷⁰ and is responsive to microglial DNA damage. ⁷¹ Recent studies report that activation of the NLRP3 and AIM2 inflammasome signaling pathways plays an important role in the neuroinflammation that drives AD pathology. ⁷¹ , ⁷² Specifically, the deletion of the AIM2 inflammasome in AD mice models promoted dendrite branching, synaptic plasticity, and improvement in spatial memory. ⁷³ Expression of NLRP3 and AIM2 was significantly higher in AD mice than in WT and decreased strikingly after UA treatment in the hippocampus of AD and ADP mice (Figure 6E,F). We also evaluated the phosphorylation of STAT3 on tyrosine 705 in both AD and ADP mice. STAT3 phosphorylation is critical for cytokine secretion and is linked to neuroinflammation in AD and to IL‐6 and TNFα. ⁷⁴ Phosphorylation of STAT3 on tyrosine 705 in the hippocampus was greatly elevated in both APP/PS1 transgenic mice ⁷⁵ and in AD patients, ⁷⁶ and also verified in our APP/PS1 mouse brains (Figure S4D,F). As expected, the ratio of phosphorylated to total STAT3 protein was elevated in the hippocampus of AD and ADP mice compared with WT mice and was significantly reduced after UA treatment (Figure 6G), suggesting that UA has beneficial effects on neuroinflammation. Several studies have provided strong evidence of crosstalk between DNA damage and inflammation. ⁷⁷ Thus we investigated the expression of γH2AX, a molecular marker of DNA damage and repair, to evaluate the effects of UA on DNA damage. As reported previously, ¹¹ , ²⁹ we showed that in the hippocampus of ADP mice, γH2AX was significantly increased compared with the WT group. Notably, after UA treatment, the AD mice showed a decrease in the protein level of γH2AX (Figure 6H,I), as well as decreased PARylation and p‐Parkin (Figure 6H and Figure S5G,H), indicating decreased DNA damage and increased DNA repair.

FIGURE 6.

Effect of UA on mitochondrial dysfunction, neuroinflammation, and DNA damage. (A) Representative western blots showing PINK1, Parkin, STAT3, p‐STAT3, NLRP3, AIM2, and Bax in the hippocampus of WT, AD, and ADP mice with/without UA; n = 3 mice per group (1 F and 2 M). (B–G) Quantitation of protein levels for indicated proteins in Figure 6A. (H–I) Representative western blots showing PAR, p‐Parkin, and γH2AX expression in the hippocampus and γH2AX quantification; n = 5 mice per group. (J–L) Representative western blots showing Ctsw and Ctsd and the quantitation of these protein levels in the hippocampus of WT, AD, and ADP mice with/without UA; n = 3 mice per group (1 F and 2 M). (M, N) Ctsz and Ctsw levels by ELISA assay. (O, P) Representative western blots showing Ctsz and the quantitation of the protein levels in the hippocampus of WT, AD, and ADP mice with/without UA, n = 5 mice per group. For H and O, WT (2 M & 3 F), AD (3 M & 2 F), AD+UA (3 M & 2 F), ADP (3 M & 2 F), and ADP+UA (2 M & 3 F). F, female, M, male. All bands were normalized to their respective loading controls β‐actin. Error bars represent the mean ± SEM. Data were analyzed by one‐way ANOVA analysis of variance with Tukey's multiple comparisons test (B–G, I, K–N, P).

To further understand the pathway and gene changes related to mitochondria and metabolism in AD and ADP mice after UA, we used the NanoString metabolism panel and found some significantly changed terms, including lysosomal degradation and NF‐κB terms, as well as several cathepsins and the Ctsz gene (Figure S6E,F). We also validated the expression of lysosomal Ctsw, Ctsd, and Ctsz in the hippocampus of AD and ADP mice models, respectively. Ctsw and Ctsd levels were lower in AD mice than in WT, and UA treatment did not notably affect their expression, except for the Ctsw levels in ADP mice (Figure 6J–M). The level of Ctsz expression was significantly elevated in AD brains and normalized after UA treatment (Figure 6N–P). Similar results were seen in APP/PS1 mice, so across multiple mouse models, we find cathepsins, and in particular Ctsz, regulated by UA.

4. DISCUSSION

Our study identifies mechanisms underlying the beneficial effects of long‐term UA treatment in AD transgenic mice models (Figure 7). APP/PS1 mice treated long term with UA took significantly less time to reach the platform, similar to WT (Figure 1B); we find that UA treatment has beneficial effects on multiple aspects of neuropathology (long‐term potentiation, Aβ, p‐tau, neuroinflammation, lysosomal function, and DNA damage) in the hippocampus and PFC regions, which play central roles in AD. As shown in the above preceding figures, key drivers of the neuroinflammatory process, including IL‐1β, NLRP3, AIM2, and STAT3, were higher in the brains of AD mice than in WT and decreased notably after UA treatment. It is notable that we are the first to present evidence that Ctsz plays an important role in UA‐induced therapeutic effects for AD (Figure 7). Thus our results are consistent with literature reporting that UA has therapeutic properties in the brains of AD mice. ¹¹ , ²¹

FIGURE 7.

Graphical abstract. From organismal, bioinformatical, and cellular aspects, the effects and mechanisms of urolithin A (UA) were investigated. We used different mouse models (including APP/PS1, 3xTgAD, and 3xTgAD/Polβ^+/−) and found that UA treatment improved behaviors, decreased inflammation, and microglia hyperactivation. By using RNA‐seq and NanoString, lysosomal pathways and genes were enriched after UA treatment. RNA and protein levels of some lysosomal cathepsins are altered by UA treatment. We propose that UA inhibits CTSZ expression, decreases Aβ and inflammation, and induces mitophagy.

We have summarized in a table comparing UA treatment for 1–2 months (which we reported previously ¹¹ ) and UA treatment for 5 months in AD mouse models (Table S1). We have compared and summarized the behavioral improvement (including cognitive function, smelling, etc.), the improvement of pathological characteristics of AD (including Aβ, tau, etc.), the neuroinflammation, the improvement of mitophagy, and other aspects of each AD mouse treated with 1‐ to 2‐month UA and 5‐month UA. We found that 5 months of UA treatment was similar to 1–2 months of UA treatment in some aspects, including that (1) UA improved the cognitive function of AD mice in cognitive behavioral experiments with MWM, object recognition, and Y‐maze; (2) UA reduced the main pathological features of AD (Aβ, tau); (3) UA reduced proinflammatory factors and the inflammasome NLRP3; and (4) mitophagy induction of UA. Of interest, 5 months of UA treatment in AD mice had novel effects in many ways: (1) UA treatment showed better olfactory ability in smelling test; (2) synaptic function was significantly improved during long‐term potentiation; (3) IBA1, GFAP, and other neuroinflammation‐related markers were significantly improved; (4) the inflammasome AIM2 and NF‐κB also had significant changes; (5) DNA damage–related markers such as PAR and γH2AX were significantly decreased; (6) increased expression levels of longevity‐related genes Sirt1 and Sirt3; (7) newly identified alterations in lysosomal associated Ctsz expression; and (8) alterations in the AD risk gene Trem2. Through this detailed comparison, we provide information for future translational research.

UA is thought to play a critical role in supporting mitochondrial function in neuronal dysfunction and neurodegeneration. ¹¹ , ²¹ Mitochondrial dysfunction has been reported widely in studies of AD patients and AD models. ⁵ Notably, we found that full‐length PINK1 and Parkin recruitment were significantly increased in AD brains, with mitochondrial damage likely leading to activated Bax (Figure 6). Damage to mitochondria results in activation of mitophagy and affects the mitochondrial outer membrane permeabilization (MOMP) that in turn triggers apoptotic cell death. Upon the collapse of the mitochondrial membrane potential (Δψm), the import and degradation of PINK1 are blocked and PINK1 accumulates on the outer mitochondrial membrane. ⁶⁷ , ⁶⁸ In addition, Bax translocates from the cytosol to the mitochondrial membrane, thereby facilitating the release of proapoptotic proteins during apoptosis. UA treatment of AD and ADP mice normalized the levels of these proteins to WT.

UA has also been shown to have a positive effect on DNA damage and repair in AD. ¹¹ DNA damage and DNA repair deficiency can lead to neuronal dysfunction in aged brains and in the brains of AD patients and transgenic mice. ²⁹ , ⁷⁸ In our study, DNA damage was identified by high levels of γH2AX in the hippocampus of brain tissue, and these decreased after treatment with UA (Figure 6H). A previous study also reported that UA suppressed DNA double‐strand breaks (DSBs) in bone marrow–derived macrophages. ⁷⁹ These findings suggest that treatment with UA may regulate DNA damage by decreasing the amount of oxidative DNA damage and mitochondrial oxidative stress or by increasing DNA repair.

UA at doses from 250 to 2000 mg in humans ²⁵ and 1–450 mg/kg in mice ⁸⁰ has been reported to be safe. UA increased muscle strength and physical performance in a 6‐min walk test in elderly humans after 4 months of supplementation. ⁸¹ Other studies reported that UA improved motor activity in the rotarod test and increased total distance traveled and average speed in the open field test in young C57BL/6J mice ⁸² and 3xTg AD mice. ⁸³ However, we noted that UA decreased the distance traveled by 3xTgAD mice in the open field test (Figure 1L), whereas the open field activity of APP/PS1 mice in our study was normal compared with age‐matched control mice (Figure S1A–D). The mice were of different ages in these two cohorts. The APP/PS1 mice were 7 months old, and the AD and ADP mice were 17 months old when testing. Another potential factor is that 3xTgAD mice express mutant tau, whereas the APP/PS1 mice do not express the transgene. In addition, another difference between the strains was sex; the APP/PS1 mice were all male, whereas the 3xTgAD were both sexes. Combined, there are several potential reasons that the strains have a different response.

Growing evidence implicates the important role of lysosomal function in AD development. Particularly, alterations in lysosomal cathepsins in the CNS contribute to the pathogenesis of neurodegenerative diseases as seen for AD, as well as synucleinopathies (PD and dementia with Lewy bodies), and HD. ⁸⁴ In our study, one of the most significant findings was that Ctsz was highly expressed in multiple AD transgenic mouse models, and its expression was normalized by UA treatment (Figures 4 and 6), suggesting a critical role of Ctsz in UA‐induced therapeutic effects for AD. Further analysis also revealed the mutual functions between UA and CTSZ. Inhibiting CTSZ activity reduced the Aβ₄₂ level in the MHC3 cells and also compromised UA's increased lysosomal function, maybe because CTSZ inhibition decreased Aβ₄₂ levels, suggesting that UA is partly dependent upon CTSZ to facilitate lysosomal activities. Over the past decade, increasing evidence has emerged implicating Ctsz (also known as cathepsin X/P) in the inflammatory processes leading to neurodegeneration. Ctsz has been shown to be highly expressed and is secreted by microglia and astrocytes in response to neuronal damage and inflammatory stimulus, both in vitro and in vivo. ⁸⁵ , ⁸⁶ Unlike other cathepsins, such as Ctsa, Ctsh, and Ctsw, Ctsz is unique in its ability to modulate NLRP3‐inflammasome activation. ⁸⁷ It is widely expressed early in CNS development and linked to interleukin 18 (IL‐18) and IL‐1β production. ⁵⁰ , ⁵¹ In particular, high levels of Ctsz have been observed in degenerating brain regions of the ALS and AD transgenic mouse models, and high levels of CTSZ in the postmortem cortex tissues of AD individuals. ⁸⁸ , ⁸⁹ In the transgenic APP/PS1 mouse model, Ctsz upregulation has been observed in microglial cells surrounding amyloid plaques, suggesting that activated microglia may have contributed to inflammation‐induced neurodegeneration by secreting Ctsz. ⁸⁹ In addition, a decreased neuroinflammatory state was observed in Ctsz knockout mice with experimental autoimmune encephalomyelitis, ⁵⁰ supporting that Ctsz may be an interesting therapeutic target in AD neuroinflammation. A large number of studies were included in one systematic review and meta‐analysis, showing significant differences in endosomal lysosomal and autophagy protein levels in the cerebrospinal fluid (CSF) of AD and healthy controls, particularly the lysosomal membrane protein CTSZ. ⁹⁰ CTSZ immunoreactivity in the brains of patients with AD has also been reported, however, in less detail. ⁸⁹ Despite this, unlike other cathepsins, only these preliminary studies have been conducted on CTSZ and AD, and no studies and mechanisms have been explored in mouse models of AD.

UA‐mediated inhibition of neuroinflammation and lysosomal dysfunction may contribute to the improvement of AD pathophysiology. Our results provide experimental evidence that long‐term UA treatment may ameliorate neuroinflammation and improve lysosomal function in AD transgenic mouse models.

AUTHOR CONTRIBUTIONS

Yujun Hou and Vilhelm A. Bohr designed the experiments. Yujun Hou, Beimeng Yang, and Yong Wei performed the animal treatment and behavior tests. Yujun Hou performed the microarray and Deborah L. Croteau analyzed the microarray data. Xixia Chu and Jae‐Hyeon Park performed the western blot, ELISA, and NanoString experiments, and Deborah L. Croteau and Jae‐Hyeon Park analyzed NanoString results. Qing Zhu, Mansoor Hussain, and Helena Borland Madsen performed the lysosomal assay and immunofluorescence assays. Zhiquan Li performed the in vitro assay. Yue Wang performed the electrophysiology experiments. Evandro F. Fang contributed to constructive discussions and editing of the manuscript. Yujun Hou, Xixia Chu, Jae‐Hyeon Park, Deborah L. Croteau, and Vilhelm A. Bohr wrote the manuscript.

CONFLICT OF INTEREST STATEMENT

E.F.F. has a CRADA arrangement with ChromaDex (USA) and is a consultant to Aladdin Healthcare Technologies (UK and Germany), the Vancouver Dementia Prevention Centre (Canada), Intellectual Labs (Norway), and MindRank AI (China). V.A.B. has a relationship with and previously had a CRADA arrangement with ChromaDex (USA). Other authors declare no conflicts of interest. Author disclosures are available in the Supporting information.

Supporting information

Supporting Information

Supporting Information

Hou Y, Chu X, Park J‐H, et al. Urolithin A improves Alzheimer's disease cognition and restores mitophagy and lysosomal functions. Alzheimer's Dement. 2024;20:4212–4233. 10.1002/alz.13847