A Xanthine Derivative With Novel Heat Shock Protein 90-Alpha Inhibitory and Senolytic Properties

Abstract

The accumulation of senescent cells contributes to aging and related diseases; therefore, discovering safe senolytic agents-compounds that selectively eliminate senescent cells-is a critical priority. Heat shock protein 90 (HSP90) inhibitors (HSP90i), traditionally investigated for cancer treatment, have shown potential as senolytic agents. However, inhibitors face formulation, toxicity, and cost challenges. To overcome these limitations, we employed a virtual screening approach combining structure-based prefiltering with a ligand-based pharmacophore model to identify novel, potentially safe HSP90 alpha isoform inhibitors exhibiting senolytic properties. This strategy identified 14 candidate molecules evaluated for senolytic activity in primary human fetal pulmonary fibroblasts. Four compounds exhibited significant HSP90i and senolytic activity, including two novel compounds, namely K4 and K5. The latter, 1-benzyl-3-(2-methylphenyl)-3,7-dihydro-1H-purine-2,6-dione, structurally related to the xanthinic family, emerged as a promising, well-tolerated senolytic agent. K5 demonstrated senolytic activity across various cellular senescence models, including human fibroblasts, mesenchymal stem cells, and breast cancer cells. It was also effective in vivo, extending lifespan in Drosophila and reducing senescence markers in geriatric mice. Additionally, the xanthinic nature of K5 implicates a multimodal action, now including the inhibition of HSP90α, that might enhance its efficacy and selectivity towards senescent cells, Senolytic index SI > 1320 for IMR90 cells, and SI > 770 for WI38 cells, underscoring its therapeutic potential. These findings advance senolytic therapy research, opening new avenues for safer interventions to combat age-related inflammaging and diseases, including cancer, and possibly extend a healthy lifespan.

Keywords: drosophila; HSP90α inhibitors; aging; senescence; senolytics; xanthine.

FIGURE 1

In vitro drug screening on IMR90 (Human fetal primary fibroblast) and target validation, K5 structure, and molecular interactions with HSP90. (A) Workflow diagram of the combined structure‐based and ligand‐based virtual screening strategy. (B) Representative images of SA‐β‐galactosidase staining on early‐passage cells (PD15, upper‐left) and advanced‐passage cells (PD19, upper‐right); 10× magnification, scale bar = 100 μm. Lower panel: The graph shows the SA‐β‐galactosidase product quantification evaluated comparing experiments on PD15‐PD17 (black bars) and PD19‐PD21 (blue bars) cells, normalizing to cells number; n = 6; mean ± SEM; data were analyzed by Mann–Whitney test; **p < 0.005. (C) The graph represents the average cell number of cells surviving treatment with the 14 in silico‐selected compounds on the two groups of cell populations PD15‐PD17 and PD19‐PD21, at 1 μM concentration, n = 3; mean ± SEM. Data were analyzed by 2way ANOVA and two‐stage linear step‐up procedure of Benjamini, Krieger, and Yekutieli; *p < 0.05; **p < 0.01, for paired comparisons; and 2way ANOVA, Dunnett's post hoc test for Vehicle comparisons among proliferating (black brackets) and senescent cells (blue brackets), ^# p < 0.05; ^## p < 0.005. (D) The graph represents HSP90‐α inhibition dose–response curves between 1 pM and 10 μM in a concentration range. (E) The table shows the extrapolated IC₅₀; n = 3; mean ± SEM. (F) Representative images of HSP90 isoform stability were tested at 46°C in the presence of HSP90 inhibitors (0.5 mM). DMSO and Caffeine were used as solvent and negative controls, while Ganetespib was used as a positive control. α‐tubulin and β‐actin were used as normalizers. (G‐H) The graphs show the normalized protein levels; n ≥ 3; mean ± SEM. 1way ANOVA and Dunnett post hoc test analyzed data; **p < 0.005. (I) 2D interaction plot of K5 with the amino acid residues within the HSP90 binding pocket. The plot was generated using the Biovia Discovery Studio using three‐dimensional structural analysis of the HSP90‐ligand complex. (J) Solid ribbon representation of HSP90 bound to K5. Key interacting residues are depicted as stick models, colored by atom type. Biovia Discovery Studio was used to visualize 3D structures. Critical interactions between HSP90 and K5, identified using geometric and chemical criteria in the Discovery Studio Visualizer, are highlighted as dashed lines. Compound K5 shows a predicted binding free energy of −43.06 kcal/mol as calculated by the MM‐GBSA scoring approach and a Glide score of −4.883 kcal/mol.

FIGURE 2

K5 senolytic effect on IMR90 and WI38 (Human fetal primary fibroblast). (A) Dose–response curves on proliferating (PD15, black line) and senescent (PD22, blue line) IMR90 cells treated with compound K5 at increasing concentrations (0.1, 0.5, 1, 5, 10, 50, 100 and 200 μM) for 48 h. (B) Dose–response curves on proliferating (PD29, black line) and senescent (PD51, blue line) WI38 cells treated with compound K5 at increasing concentrations (0.1, 0.5, 1, 5, 10, 50, 100 and 200 μM) for 48 h. Vehicle‐treated cells (DMSO) were used as solvent control. All cells were counted using Trypan blue reagent and EVAplus cell counter; data are presented as mean ± SEM, n = 3. (C, D) Phase contrast (PHC, left) and SA‐β‐galactosidase staining (center) representative images of senescent IMR90 (PD22, C) and WI38 cells (PD51, D), treated with K5 at 1 μM (yellow boxes), 10 μM (orange boxes), and 100 μM (red boxes), 17‐DMAG at 5 μM as positive control (purple boxes), and Vehicle as solvent control. Images were acquired at 10× magnification, scale bar = 100 μm. The 3D plots on the right show the densitometry of the blue colorimetric signal of the SA‐β‐galactosidase product, analyzed by ImageJ software. (E) Immunofluorescence representative images showing the Cytochrome C localization (green signal) in senescent IMR90 cells (PD21) after 48 h treatment with K5 25 μM, Dasatinib 1 μM as positive control, and Vehicle as solvent control (DMSO). DAPI (blue signal) and MitoTracker orange (red signal) were used to visualize nuclei and mitochondria, respectively. Images were acquired at 40× magnification, scale bar = 50 μm. The densitometry and the colocalization scatter plots were analyzed by ImageJ using the 3D and CoFinder plugins, showing a Pearson's Rr (PRr) = 0.769 and Overlap R (OR) = 0.913 for cells treated with Vehicle, PRr = 0.109 and OR = 0.454 after K5 treatment, and PRr = 0.643 and OR = 0.639 for Dasatinib treated cells.

FIGURE 3

K5 senolytic effect on Tamoxifen‐induced senescent MCF7 breast cancer cells. (A, B) Representative images of SA‐β‐gal assay on MCF7 cells were obtained using the phase microscopy EVOS XL Core at 10× magnification and SA‐β‐gal assay performed on MCF7 cells after treatment with TAM (10 μM, 96 h) + K5 (10 μM, orange) or Vehicle (DMSO, gray) for 24 h, 48 h, 72 h, and 96 h (scale bar = 100 μm). The NT condition was treated with the TAM vehicle (EtOH 75%). Data plotted as the percentage of senescent cells (blue signal) over the total number of cells are represented as mean ± SEM of 5 independent experiments. Statistical significance was determined using a parametric paired two‐tailed Student's t‐test. *p < 0.05; **p < 0.01; ***p < 0.001. (C) Tamoxifen‐treated MCF7 cell mortality curves after treatment with K5 (10 μM, orange) or Vehicle (DMSO, gray) were performed using the Incucyte live‐cell Analysis system (Sartorius). Data plotted as the number of cells/well normalized to t0 (time of starting treatment) are represented as mean ± SEM of three independent experiments, each performed in quadruplicate. (D, E) Ki67 expression (dark brown) was evaluated by immunohistochemistry (IHC) upon Tamoxifen (10 μM, 96 h) + K5 (10μM) or Vehicle treatment (DMSO, 48 h and 96 h) in MCF7 cells—representative images obtained using the phase microscopy Nikon ECLIPSE Ei R at 40× magnification, scale bar = 50 μm. Data, plotted as the number of Ki67 positive cells (dark brown)/number of total cells expressed as a percentage, are represented as mean ± SEM of three independent experiments. Statistical significance was determined using a parametric paired two‐tailed Student's t‐test. *p < 0.05; **p < 0.01; ***p < 0.001. (F) Tamoxifen treated MCF7 cells viability curve after treatment with K5 (10 μM, orange), compared to Vehicle (DMSO, gray) performed using the Incucyte live‐cell Analysis system (Sartorius). Data, plotted as the number of cells/well normalized to t0 (time of starting treatment), are expressed as mean ± SEM of three independent experiments, each performed in quadruplicate. Statistical significance was determined using a parametric paired two‐tailed Student's t‐test; *p < 0.05. (G, H) Annexin V (blue signal) assay performed using the Incucyte live‐cell Analysis system (Sartorius) upon Tamoxifen (10 μM; 96 h) + K5 (10 μM) or Vehicle treatment (DMSO, over 96 h) in MCF7 cells—representative images obtained using the Incucyte (Sartorius, G), scale bar = 400 μm. (H) Data, plotted as the cells mean signal intensity (Annexin V positive cells, blue)/number of total cells, are represented as mean ± SEM of three independent experiments, each performed in quadruplicate. Statistical significance was determined using a parametric paired two‐tailed Student's t‐test. *p < 0.05. (I) Dose–response curves on proliferating (black line) and senescent (TAM, 10 μM, 96 h, blue line) MCF7 cells treated with compound K5 at increasing concentrations (2.5, 5, 10, 50, 100, 250, and 500 μM) for 96 h. Vehicle‐treated cells (DMSO) were used as solvent control—evaluation on proliferating or senescent MCF7 cells. The experiments were performed using the Incucyte live‐cell Analysis system (Sartorius). Data was plotted as a percentage of cell reduction vs control normalized to t0 (time of starting treatment).

FIGURE 4

Effect of K5 on Drosophila melanogaster lifespan and gene expression of senescence‐related genes. (A) The panel shows the initial experiment testing two concentrations of K5 (10 μg/mL and 100 μg/mL) compared to control (CTRL) using 100 flies per group. Both concentrations showed improved survival, with a statistically significant effect (p = 0.034). Interestingly, the lower dose (10 μg/mL) appeared to perform better than the higher dose (100 μg/mL). (B) The graph represents the same data of panel A without the 100 μg/mL K5 group. Statistics are computed only by comparing the 10 μg/mL K5 group vs controls (CTRL). Here, the survival benefit becomes more evident, with K5‐treated flies showing consistently better survival throughout their lifespan (p = 0.012). The survival curves diverge, with K5‐treated flies maintaining higher survival rates than controls. (C) The panel presents an exciting finding in a larger population (n = 884, 445 males and 439 females) where K5 treatment (10 μg/mL) was started later in life (at day 20). Even with this delayed treatment start, K5 showed significant beneficial effects (p = 0.003). The data show that K5‐treated flies lived 50.69 days (±0.79 SE) on average, a 6.63% increase compared to control flies (47.54 days ± 0.74 SE). (D) The graph shows the relative expression levels (2^−ΔCt) of Dap, Upd2, and Upd3 in the K5‐treated group (10 μg/mL, red) and the control group (cyan). Dap expression was similar between groups, with no significant difference. In contrast, Upd2 and Upd3 expression levels were significantly lower in the K5‐treated group than in controls (*p < 0.05). Data represent means (±SD) from three independent experiments performed in triplicate, with individual means shown as single dots. (E, F) The graphs show the survival data from panel C, restricted to females (E, p = 0.002) and males (F, p = 0.044). These findings suggest that K5 extends lifespan when administered throughout life, reduces senescence markers, and provides benefits when treatment begins in mature flies. Notably, the effects were significant in both females and males, highlighting its potential therapeutic value even when intervention starts later in life.

FIGURE 5

Live imaging and cellular senescence markers of p16‐3MR mice from the pilot study. (A) Average ventral (left) and dorsal (right) Luminescence of p16‐3MR mice treated with K5 (n = 9) compared to control p16‐3MR mice (n = 8) of the same age. All mice were treated at the age of 25 months. Data are expressed as mean (95% CI). *p < 0.05 by Wilcoxon test. (B) Representative images of p16‐3MR mice treated with K5 and untreated controls. (C) Expression of p21 and p53 and percentage of cells positive for SA‐β‐gal in ear biopsies of mice treated with K5 (n = 5–8). Data are shown as mean (95% CI). *p < 0.05 by Wilcoxon test. (D) Circulating levels of IL‐6 and IL‐1β in mice treated with K5 (n = 9; 5F and 4 M). Data are expressed as mean (95% CI). *p < 0.05 by Wilcoxon test.

FIGURE 6

Clinical and functional status of geriatric mice (> 30 months) treated with K5. (A) Glucose, Creatinine, GOT, and GPT serum concentrations in treated mice compared to the control group. Data are expressed as mean (95% CI). (B) Body weight of male and female mice. Data are expressed as mean (95% CI). (C) Frailty Index, Physical function, and Vitality Score adjusted for baseline levels. Data are expressed as mean (95% CI). *p < 0.05 by Mann–Whitney U test.

FIGURE 7

Cellular senescence markers in geriatric mice (> 30 months) treated with K5. (A) Circulating IL‐6 and IL‐1b levels in mice treated with K5 versus controls. Data are expressed as mean (95% CI). *p < 0.05 by General Linear Model (GLM) with a univariate approach, adjusting for sex as a covariate. (B) Multiorgan SA‐β‐gal assay showed as the mean % of SA‐β‐galactosidase positive cells (% SA‐β‐gal⁺ cells) of K5 (n = 20) and control (n = 20) mice analyzed in kidney, liver, lung, cortex, heart, quadriceps, cerebellum, pancreas, gastrocnemius, and hippocampus. Data are expressed as mean (95% CI). *p < 0.05 by General Linear Model (GLM) with a univariate approach, adjusting for sex as a covariate.