Tomatidine may improve thinking and reduce aging cells in old mice

From the small screen, tomatidine, an aglycone of tomatine naturally found in tomatoes, was identified as a promising candidate with both senolytic and senomorphic activity in a concentration-dependent manner (Fig. EV1A–D). Tomatidine reduced the percent of C₁₂FDG-positive cells (senescent Ercc1^−/− MEFs) with an EC₅₀ value of 1.241 µM. At low concentrations, there was no reduction in the cell number of Ercc1^−/− MEFs, consistent with a senomorphic effect, whereas at high doses there was a reduction in the total number of cells, consistent with a senolytic effect (EC₅₀ = 16.73 µM). Tomatidine had only a marginal effect on non-senescent WT MEFs at lower concentrations (EC₅₀ = 41.52 µM), yielding a good selectivity index (SI) of 33.5 (Fig. EV1C). Given the known safety of tomatidine as a naturally occurring compound (Dyle et al, 2014), its oral bioavailability, and prior evidence linking it to improved mitochondrial function, inhibited skeletal muscle atrophy and reduced inflammation in aging models, we further examined the senotherapeutic effects of tomatidine (Ahsan et al, 2020; Ebert et al, 2015; Fang et al, 2017).

We next tested tomatidine in two distinct senescent human cell types: lung fibroblasts (IMR90) and human brain microvascular endothelial cells (HBMEC), both induced into senescence by X-ray irradiation. In neither cell type did tomatidine exhibit senolytic activity; however, it significantly reduced the expression of SASP-related genes without altering p21^CIP1 or p16^INK4a levels, consistent with a senomorphic effect (Fig. EV1E−I). Treatment of senescent HBMECs showed no significant reduction in p21^CIP1 and p16^INK4a when treated with tomatidine at day 10 for 48 h (Fig. EV1J,K), again confirming that tomatidine has no senolytic activity on certain senescent cell types.

Cognitive performance was first evaluated using the Y maze, where increased time spent in the novel arm reflects enhanced spatial working memory. Tomatidine-treated mice demonstrated significant improvements in this task (Fig. 1D,E). Similarly, in the Stone T maze, a test of spatial learning and memory, tomatidine-treated mice made significantly fewer errors, reaching performance levels comparable to young controls (Figs. 1F,G and EV2D,G–J). While the overall time required to complete the maze did not differ between groups (Fig. EV2E), treated mice displayed enhanced learning across trials. Specifically, they showed significant reductions in errors between trials 1–3 vs. 7–9 and 4–6 vs. 7–9, whereas untreated mice improved only between trials 1–3 vs. 7–9 (Fig. EV2D).

These findings identify tomatidine as a promising senotherapeutic compound with both in vitro and in vivo efficacy. Following these observations, we sought to examine how it impacts senescence at the tissue level.

To assess neuroinflammation, we analyzed Iba1⁺ microglia. In aged mice, tomatidine reduced both microglial number and soma size in the hippocampus (Fig. 3D,E), consistent with attenuated microglial activation. Comparisons with young mice showed the expected age-related increase in microglial number and soma size in the hippocampus, and an age-related increase in microglial number, but not soma size, in the cortex (Fig. EV4C–F). Tomatidine partially reversed these age-associated changes. In addition, cytokine profiling of hippocampal homogenates revealed broad reductions in pro-inflammatory factors in tomatidine-treated mice (Fig. 3F,G). Together, these findings indicate that tomatidine alleviates neuronal senescence and neuroinflammation in the aged brain.

Given these improvements in neuronal senescence and inflammation, we next asked whether tomatidine also impacts brain endothelial cells (BECs), which form the structural and functional basis of the blood–brain barrier (BBB).

BBB dysfunction is a well-established feature of aging and has been linked to increased permeability and altered expression of tight junction proteins (Andjelkovic et al, 2023; Zhao et al, 2020). Recent work has shown that aging-associated BBB disruption is, at least in part, mediated by BEC senescence, suggesting that senescent endothelial cells contribute to the transition from a young to an aged BBB phenotype (Novo et al, 2024). Based on this, we sought to determine whether tomatidine could modulate BEC senescence in aged mice.

Altogether, these results identify tomatidine as a compound that counteracts senescence-associated decline and supports brain health during aging.

Experimental procedures were approved by the Institutional Animal Care and Use Committee (IACUC) at Mayo Clinic (protocol A00005445). C57BL/6 wild-type mice (both sexes) were purchased from Jackson Laboratory (Bar Harbor, ME, USA) at 20 months of age and housed in a pathogen-free facility under controlled environmental conditions, including a temperature range of 23–24 °C and a 12-h light/dark cycle. Mice were maintained in static, autoclaved HEPA-ventilated microisolator cages (27 × 6.5 × 15.5 cm) containing autoclaved Enrich-o’Cobs bedding. Bedding and cages were changed biweekly in class II biosafety cabinets. Routine quarterly pathogen testing consistently yielded negative results. At 21 months of age, mice were randomly assigned to one of two dietary groups. Animals received ad libitum access to water and either standard mouse chow (Lab Diet 5053, St. Louis, MO) or chow supplemented with 0.05% tomatidine (Diets Inc., New Jersey, USA). Tomatidine (BML-GR335-0000; ≥90% purity by thin-layer chromatography, TLC) was purchased from Enzo Life Sciences (Farmingdale, NY, USA) and incorporated into the diet by Research Diets Inc. (New Brunswick, NJ, USA). The dietary intervention was maintained for 3 months. For comparison, we included a separate cohort of 4-month-old young mice (C57BL/6) of both sexes, also obtained from Jackson Laboratory, which served as untreated controls. At the end of the treatment period, mice were euthanized, and organs were harvested for downstream analyses. Mouse tissues were either snap-frozen in liquid nitrogen for biochemical analysis or fixed in 4% paraformaldehyde (PFA) for 24 h. Fixed samples were then processed, embedded in paraffin, and sectioned at 3-μm thickness for further analysis.

Senescence was evaluated based on SA-β-gal activity using a C12FDG staining assay. Senescent Ercc1^−/− MEFs were generated by passaging the cells three times at 20% O₂ and then seeded at 3000 cells per well in black-wall, clear-bottom 96-well plates at least 16 h prior to treatment. Following the addition of compounds, cells were incubated for 48 h at 20% O₂. After removing the medium, cells were incubated with 100 nM Bafilomycin A1 in culture medium for 60 min to induce lysosomal alkalinization, followed by 20 μM C12FDG (7188, Setareh Biotech, USA) for 2 h and counterstaining with 2 μg/mL Hoechst 33342 (H1399, Thermo Fisher Scientific, MA, USA) for 15 min. Cells were then washed with PBS, fixed in 2% paraformaldehyde for 15 min, and imaged with six fields per well using a Cytation 1 high-content fluorescence imaging system (BioTek, VT, USA).

Spatial and working memory assessments in mice were conducted using a Y maze setup. The Y maze apparatus consists of three identical arms with elevated walls. Mice were habituated to the testing room for at least one hour prior to the experiment. During the training phase, one arm (the novel arm) was blocked, and mice explored the other two arms for 10 min. Following a 30-min rest period, the test phase was conducted, during which all three arms were accessible for 5 min. Behavior was recorded and analyzed using EthoVision XT (Noldus, Wageningen, Netherlands). Cognitive performance was evaluated by measuring spontaneous alternation, specifically the number of entries into the novel arm and the total time spent in it.

Cognitive performance was assessed using a water-motivated version of the Stone T maze, custom-built by the Mayo Clinic workshop, which is sensitive to age-related deficits in learning and memory. The apparatus consisted of an acrylic-roofed maze positioned in a shallow steel pan filled with water (~3 cm), motivating mice to walk and to navigate toward a dry goal box. On Day 1, mice underwent straight-run training to acclimate to the escape paradigm. Each mouse was placed in the start box and allowed to traverse a straight maze to reach the goal box. Upon entry, mice were gently dried with a towel and placed in a heated holding cage. The testing room was maintained at a comfortable ambient temperature, and the holding cage was partially warmed with a heat lamp, allowing self-regulated exposure to heat. Each mouse completed three straight-run trials per session, with three sessions separated by 15-min intervals, totaling nine training trials. Maze testing was conducted the following day. After one hour of habituation in the testing room, each mouse was placed in the start box and allowed to explore the maze for up to 6 min. Latency to reach the goal box and the number of errors defined as full-body entries into incorrect arms, including each distinct path at T-junction dead ends were recorded. Inter-trial intervals of 15 min were maintained. After each trial, mice were dried with a towel and returned to the heated holding cage. Trials in which a mouse failed to reach the goal box within 5 min were marked as unsuccessful. Mice exhibiting three consecutive unsuccessful trials or signs of impaired thermoregulation (e.g., excessive shaking) were excluded from further testing. Each mouse completed a total of nine acquisition trials.

Anxiety-like behavior and locomotor activity were assessed using the elevated plus maze (EPM) test. The apparatus consisted of two open arms (25 × 5 cm each) and two closed arms (25 × 5 cm each), arranged at right angles to a central platform (5 × 5 cm) and elevated 40 cm above the floor. To minimize environmental stress, mice were acclimatized to the testing room for at least 1 h prior to the experiment. Each mouse was placed individually at the center of the maze, facing an open arm, and allowed to explore freely for 5 min. Behavior was recorded using a video camera and analyzed with EthoVision XT tracking software (Noldus Information Technology, Wageningen, Netherlands). Anxiety-like behavior was evaluated by quantifying the number of entries into open arms and the total time spent in them.

Motor coordination and maximal walking speed were assessed using an accelerating RotaRod (Ugo Basile, RotaRod 47650). Mice were trained for three consecutive days prior to testing. Training sessions consisted of mice remaining on the rotating rod at constant speeds of 4, 6, and 8 revolutions per minute (rpm) for up to 300 s on days 1, 2, and 3, respectively. On the test day, the rod accelerated linearly from 4 to 40 rpm over 300 s. The latency to fall and the corresponding speed at the time of fall were recorded. Each mouse completed three test trials, and the average performance across trials was used for analysis.

Balance and motor coordination were assessed using the Pole Test. Mice were placed at the center of a horizontal pole (60 cm in length), elevated 50 cm above ground. To ensure animal safety, a box filled with soft padding was positioned beneath the apparatus to cushion potential falls. During each trial, mice were allowed to move freely in either direction along the pole for up to 60 s. A trial was considered successful if the mouse remained on the pole for the entire duration without falling. Each mouse completed five trials, with 30-s rest intervals between trials. Time spent on the pole during each trial was recorded as a measure of motor coordination and balance.

Forelimb grip strength was measured using a grip strength meter equipped with a computer-integrated force transducer. Mice were allowed to grasp a horizontal metal bar with their forelimbs and were then gently pulled backward until they released the bar. The peak force exerted during each attempt was recorded. Each mouse completed three consecutive trials, and the average peak force was used for analysis.

Frailty was evaluated using a 27-parameter clinical frailty index, as previously described by (Whitehead et al, 2013). Each parameter was scored on a three-point scale: 0 for absent, 0.5 for mild, and 1 for severe phenotype. Assessed parameters included evaluations of the integument, musculoskeletal system, vestibulocochlear/auditory function, ocular and nasal systems, digestive system, urogenital system, respiratory system, signs of discomfort, as well as measurements of body weight (g) and body surface temperature (°C). This index provides a non-invasive, quantitative measure of overall health status in aging mice. The frailty index was calculated as the average score across all parameters assessed. We measured body weight, body temperature, fore limb strength but these measures are not included in the combinded frailty index.

Human IMR90 lung fibroblasts were obtained from American Type Culture Collection (ATCC) and cultured in EMEM medium with 10% FBS and pen/strep antibiotics. Cells were obtained from an authenticated commercial source and were used without further independent authentication. Mycoplasma testing is performed regularly. To induce senescence, cells were treated with 20 Gy X-ray irradiation.

Human brain endothelial cells (HBECs, #1000) were cultured in Endothelial Cell Medium (ScienCell, #1001) supplemented with FBS (#0025), endothelial cell growth supplement (#1052) and penicillin/streptomycin (#0503). Cells were obtained from an authenticated commercial source and were used without further independent authentication. Mycoplasma testing is performed regularly. Cells were maintained at 37 °C in a humidified incubator with 3% O₂ and 5% CO₂. To induce senescence, cells were treated with 10 Gy X-ray irradiation.

Total RNA was isolated from homogenized tissue biopsies or cells using TRI Reagent (Sigma-Aldrich) or the RNeasy Mini Kit (Qiagen, 74106), following the manufacturer’s instructions. Complementary DNA (cDNA) was synthesized using the High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher, 4368814) according to the manufacturer’s protocol. Quantitative real-time PCR (RT-qPCR) was performed using one of the following reagent kits: Power SYBR® Green PCR Master Mix (Invitrogen, 4367659) on a C100™ Thermal Cycler (Bio-Rad), PerfeCTa ToughMix (QuantaBio, 95112–250) on a CFX96™ Real-Time System (Bio-Rad), or Brilliant III Ultra-Fast SYBR Green qPCR Master Mix (Agilent Technologies) on a Bio-Rad Real-Time PCR System. Data acquisition and analysis were done using Bio-Rad CFX Manager software. Predesigned primers and probes from IDT PrimeTime.

Previously prepared hippocampal homogenates were used for protein analysis. Protein concentrations were determined using the Bio-Rad protein assay (Bio-Rad, reagent A, 500-0113; reagent B, 500-0114; reagent C, 500-0115). Equal amounts of protein from each sample were resolved on Tris-glycine gels, and samples were then blotted onto a 0.45 μm polyvinylidene difluoride (PVDF) membrane (Millipore) using Trans-Blot SD Semi-Dry Transfer Cells (Bio-Rad). Membranes were blocked with 5% milk powder in TBS-T (TBS with 0.05% Tween-20 in PBS) at room temperature for 1 h and then incubated with primary antibodies in 1x TBS-T with 5% BSA overnight. The primary antibodies used were rabbit anti-claudin 5 (1:1000; 34-1600, Invitrogen), rabbit anti-occludin (1:1000; AB216327↗, Abcam), goat anti-ve-cadherin (1:1000; #AF1002, R&D systems), and rabbit anti-actin (1:1000; A2066. Thermo Fisher). After washing in TBS-T, the membranes were incubated with a peroxidase-conjugated secondary antibody in a dilution of 1:5000 for 1 h at room temperature. The membranes were then incubated with either Clarity ECL Western Blot Substrate (Bio-Rad, 170–5060) or the KwikQuant Western blot detection kit (Kindle Bioscience, R1100) according to the manufacturer’s instructions and visualized using iBright 1500 (Invitrogen). The optical density of immunoreactive bands was quantified using Quantity One software (Bio-Rad, California, USA). Protein expression levels were normalized to β-Actin and reported as relative values compared to the control group.

Fixed single-cell suspensions were prepared from flash-frozen mouse brain tissue using the Chromium Next GEM Single-Cell Fixed RNA Preparation Kit (10x Genomics, CG000553↗, Rev B), following the manufacturer’s protocol. Tissue dissociation was performed with the gentleMACS Octo Dissociator (Miltenyi Biotec). Library construction was carried out using the Chromium Fixed RNA Profiling Kit (10x Genomics, CG000477↗, Rev C) in a single-plex format. Probe hybridization was conducted for 20 h at 42 °C, and indexing PCR was performed for 12 cycles. Prepared libraries were sequenced on the AVITI Sequencing Platform (Element Biosciences) using the AVITI 2 × 75 High Output Cloudbreak Freestyle Kit (Cat. No. 860-00015, Element Biosciences.

Sequencing data were aligned to the murine reference genome mm10. Data with at least 500 unique molecular identifiers (UMIs), log10 genes per UMI >0.8, >250 genes per cell, and a mitochondrial ratio of less than 20% were extracted, normalized, and integrated using the Seurat package v5.2.1 in R4.4.2. After quality control and integration, we performed a modularity-optimized Louvain clustering, leading to 13 distinct clusters. Subsequently, we performed the labeling for these 13 clusters manually with established key marker genes. The dotplot was designed using the Seurat package in accordance with the ggplot2 v3.5.2 package. The feature plots were created with the Seurat package, and the percentage of p16+ and p21+ cells was calculated according to any normalized count within each cell.

Initial QC was performed using CellRanger summary metrics for both the control and treated samples. The control library contained 15,912 cells, and the treatment library contained 15,185 cells, with high-confidence read mapping in both datasets (ctrl: 84.68%, treat: 84.10%). Median per-cell sequencing depth and complexity were similar between samples: The control cells showed a median read per cell of 13,191 (median UMI/cell 7557, median detected genes/cell 3498), while the treatment cells showed a median read per cell of 13,347 (median UMI/cell 7.891, median detected genes/cell 3690).

Cells with low library complexity were removed using standard thresholds (nFeature_RNA <200 or > 6000; nCount_RNA <500). Cells with >10–15% mitochondrial transcripts were discarded. Doublets were removed using DoubletFinder with sample-specific expected doublet rates (1–2% for ~15k cells). Only high-confidence singlets were retained. After QC filtering, both samples retained a comparable number of high-quality cells suitable for downstream integration and analysis, a batch correction was performed with Seurat v5 (RPCA-based) removing technical effects. After integration, UMAP was inspected to confirm successful mitigation of batch effects.

Cytokine levels in hippocampal homogenates were quantified using a multiplex immunoassay platform. Tissue samples were homogenized in lysis buffer containing protease inhibitors and centrifuged to remove debris. Supernatants were collected, and total protein concentration was determined using the Bio-Rad protein assay. Samples were then submitted to Eve Technologies Corporation (Calgary, Alberta, Canada) for analysis using the Mouse Cytokine/Chemokine 18-Plex Discovery Assay® (MDHSTC18) platform. The assay was performed according to the company’s standard protocols. Cytokine concentrations were calculated from standard curves and are expressed in pg/mL. All values were normalized to total protein content and reported relative to control groups.

Formalin-fixed, paraffin-embedded (FFPE) tissue sections (3 μm) were deparaffinized in HistoClear and rehydrated through a graded ethanol series: 100% ethanol, 90% ethanol, and 70% ethanol, followed by two washes in distilled water. Antigen retrieval was performed by boiling slides in citrate buffer (pH 6.0; Agilent-Dako, S236984) for 10 min, followed by cooling at room temperature for 20 min and two PBS washes (5 min each). Sections were blocked in 1% bovine serum albumin (BSA), 1:60 normal goat serum or fetal bovine serum in PBS for 30 min at room temperature, then incubated overnight at 4 °C with the following primary antibodies: rabbit anti-Iba1, rabbit anti-GLUT1 and mouse anti-ZO-1. After washing with PBS, sections were incubated with species-specific secondary antibodies for 1 h at room temperature. Following three PBS washes, slides were mounted with ProLong™ Gold Antifade Mountant containing DAPI (Invitrogen) for nuclear counterstaining. For all immunohistochemical analyses, including RNA-ISH and Immuno-FISH, quantification parameters depended on the specific marker assessed. For tissue-based senescence markers (e.g., p16, p21, and TAF), a minimum of 100 cells per animal were analyzed. The number of fields examined was kept consistent across animals and experimental groups for each staining. The total number of images acquired per animal depended on the magnification used during imaging (e.g., 20× or 40× for general tissue analysis and 63× for TAF quantification).

RNA in situ hybridization was performed using the RNAscope® 2.5 HD Reagent Kit-RED (Advanced Cell Diagnostics, ACD) according to the manufacturer’s protocol. Formalin-fixed, paraffin-embedded (FFPE) tissue sections were deparaffinized in HistoClear (2 × 5 min), rehydrated in 100% ethanol (2 × 1 min), and air-dried. Sections were incubated with hydrogen peroxide (H₂O₂) for 10 min at room temperature, followed by two washes in distilled water. Antigen retrieval was performed by boiling slides in 1X RNAscope Target Retrieval solution for 15 min. After rinsing in distilled water and 100% ethanol, slides were air-dried and incubated with Protease Plus reagent for 30 min at 40 °C. Tissue sections were then hybridized with target-specific probes against p21 (ACD #408551) and p16 (ACD #411011) for 2 h at 40 °C. Signal amplification was carried out sequentially using AMP1 (30 min at 40 °C), AMP2 (15 min at 40 °C), AMP3 (30 min at 40 °C), AMP4 (15 min at 40 °C), AMP5 (30 min at room temperature), and AMP6 (15 min at room temperature), with ACD wash buffer (WB) used for intermediate washes between steps. Chromogenic detection was performed using the RNAscope 2.5 HD Reagent Kit-RED. Following five final washes in distilled water, slides were counterstained and mounted using ProLong™ Gold Antifade Mountant with DAPI (Invitrogen). RNA-ISH quantification was performed by manually scoring positive cells relative to the total number of cells in each field. Nuclei were first identified and counted based on DAPI staining to determine the total cell number. Cells were classified as RNA-ISH–positive if they exhibited one or more foci corresponding to p16 or p21 transcripts. The percentage of RNA-ISH–positive cells was calculated as the ratio of positive cells to total DAPI-positive cells.

Formalin-fixed, paraffin-embedded (FFPE) tissue sections were deparaffinized in 100% HistoClear and rehydrated through a graded ethanol series (100, 90, and 70%; 2 × 5 min each), followed by two 5-min rinses in distilled water. Antigen retrieval was performed by immersing slides in 0.01 M citrate buffer (pH 6.0) and heating to boiling for 10 min. After cooling to room temperature, sections were rinsed in distilled water for 5 min. Tissue sections were blocked using normal goat serum (1:60 in PBS with 1% BSA) for 30 min at room temperature, followed by additional blocking with an Avidin/Biotin Blocking Kit (Vector Laboratories, Burlingame, CA) for 15 min per step. Sections were then incubated overnight at 4 °C with a rabbit monoclonal anti-γH2AX antibody (1:200; Cell Signaling Technology, #9718. Following three PBS washes, sections were incubated with a biotinylated goat anti-rabbit secondary antibody (1:200; Vector Laboratories, PK-6101) for 30 min at room temperature. After three more PBS washes, fluorescein-conjugated Streptavidin-Cy5 (1:500; Vector Laboratories, A-2011) was applied for 30 min at room temperature. Slides were washed again in PBS (3 × 5 min), then post-fixed with 4% paraformaldehyde in PBS for 20 min to stabilize antibody complexes. Sections were subsequently dehydrated in a cold ethanol gradient (70, 90, and 100%; 3 min each) and air-dried. For telomere-specific peptide nucleic acid fluorescence in situ hybridization (PNA-FISH), 10 μL of hybridization solution, containing 70% deionized formamide, 20 mM MgCl₂, 1 M Tris (pH 7.2), 5% blocking reagent, and 2.5 μg/mL Cy3-labeled (CCCTAA) telomere-specific PNA probe (PANAGENE), was applied to each section. Slides were denatured at 80 °C for 10 min and hybridized for 2 h at room temperature in the dark. Post-hybridization washes included 70% formamide in 2× SSC (10 min), followed by 2× SSC (10 min) and PBS (10 min). Slides were mounted with ProLong™ Gold Antifade Mountant with DAPI (Invitrogen) and imaged using high-resolution z-stacked fluorescence microscopy with a 63× objective lens. For telomere-associated foci (TAF) quantification, z-stack images were acquired to visualize DNA damage foci in three dimensions (as we have done before (Hewitt et al, 2012)). TAF were manually quantified by first identifying γ-H2AX foci within each nucleus, followed by detection of telomeric signals using a telomere-specific probe. A TAF was identified when a γ-H2AX foci colocalized with a telomeric signal throughout the z-stack. The total number of TAF per nucleus was subsequently recorded.

Microscopic imaging was performed using a Leica DFC7000GT inverted microscope.

All results are expressed as mean ± standard error of the mean (SEM). Sample sizes were determined based on prior experience with similar experimental paradigms and on previously published studies, without formal power calculation. Animals were randomly assigned to experimental groups where applicable. Investigators were blinded to group allocation during data collection and analysis when feasible. Inclusion criteria were defined prior to experimentation, and no data were excluded unless technical issues (e.g., failed assays or sample loss) occurred. All statistical analyses, including the tests used, exact n numbers, and p values, are reported in the figure legends. Data distribution was assessed prior to selecting appropriate parametric or non-parametric statistical tests. Statistical analyses and graph generation were performed using GraphPad Prism version 10 (GraphPad Software, San Diego, CA). Differences between groups were assessed using unpaired two-tailed Student’s t-tests, one-way or two-way analysis of variance (ANOVA), followed by Tukey’s multiple comparisons test. A p value <0.05 was considered statistically significant. Significance levels are represented as follows: p < 0.05 (*), p < 0.01 (**), p < 0.001 (***), and p < 0.0001 (****) as indicated in the respective figure legends.

Schematic illustrations for experimental design and organ-specific diagrams (liver, skin, and brain) in Figs. 1, 2, and 4, as well as Figs. EV1 and EV4 were created with BioRender.com.