Hypothesis

The hexosamine biosynthetic pathway (HBP) sets a bistable switch in O-GlcNAc cycling that determines whether a hormetic stimulus triggers a transient threat‑response or drives a sustained rejuvenation program. When HBP flux is moderate, O-GlcNAc transferase (OGT) activity spikes just enough to activate stress-responsive transcription factors (e.g., ATF4, Nrf2) and chaperones, producing the classic hormetic signature of improved memory or insulin sensitivity. If flux remains elevated beyond a threshold, O-GlcNAc transferase outpaces O-GlcNAcase (OGA), causing chronic hyper-O-GlcNAcylation of proteostasis regulators, which locks cells into a maladaptive state that mimics aging despite superficial benefits.

Mechanistic Basis

HBP flux generates UDP-GlcNAc, the donor for O-GlcNAcylation. Recent work shows that HBP activation, not hyperglycemia, drives glucotoxicity and protein aggregation in diabetic models study. Persistent O-GlcNAcylation sustains endothelial dysfunction even after glucose normalization study. In neurons, altered O-GlcNAc cycling shifts tau and amyloid-beta aggregation study. Importantly, HBP activation can improve memory while simultaneously promoting insulin resistance study, suggesting that the observed "benefit" is a compensatory threat response rather than true longevity.

We propose that the cell interprets any rise in UDP-Glc

Hypothesis: BPC-157 Acts as an Extracellular Modulator of Mitochondrial ATP Synthase β‑Subunit to Amplify NO‑Dependent Angiogenic Signaling

Core idea: BPC-157 does not require a classic GPCR; instead, it binds the β‑subunit of F₁F₀‑ATP synthase exposed on the outer membrane of endothelial cells, enhancing its catalytic efficiency. This interaction lowers mitochondrial ROS, shifts the redox balance toward NADPH oxidase inhibition, and permits sustained eNOS coupling and NO production, thereby driving angiogenesis and tissue repair.

Testable predictions

1. Binding assay – Recombinant BPC-157 will pull down ATP synthase β‑subunit from endothelial lysates, an interaction blocked by oligomycin or β‑subunit‑specific antibodies.
2. Functional read‑out – siRNA knock‑down of the ATP5B gene (encoding the β‑subunit) will abolish BPC-157‑induced Akt phosphorylation, eNOS Ser1177 phosphorylation, and NO release in HUVECs.
3. ROS modulation – BPC-157 treatment will decrease mitochondrial superoxide (MitoSOX signal) in wild‑type cells but not in ATP5B‑deficient cells.
4. Angiogenesis – In a matrigel tube‑formation assay, BPC-157‑stimulated tube length will be rescued by exogenous NO donor but not by NOS inhibitor L‑NAME, and will be lost when ATP5B is knocked down.
5. In vivo validation – Mice with endothelial‑specific ATP5B heterozygous deletion will show impaired wound‑closure acceleration by BPC-157 compared with wild‑type littermates, despite identical peptide dosing.

Mechanistic rationale Both BPC-157 and TB‑4 converge on PI3K/Akt/eNOS and a

Hypothesis

Age‑related decline of ZMPSTE24 leads to accumulation of farnesylated prelamin A that drives aberrant phase separation of the nuclear lamina, creating microdomains that sequester multiple DDR kinases (CK2α, ATM, DNA‑PKcs) and suppress their activity, thereby causing replication stress and R‑loop‑mediated genome instability.

Mechanistic Rationale

1. Farnesylated prelamin A is more hydrophobic and prone to self‑assembly, a property known to promote liquid‑liquid phase separation (LLPS) of nuclear envelope proteins.
2. The lamina already functions as a signaling scaffold that tethers kinases; LLPS would increase the local concentration and residency time of these kinases within prelamin A‑rich condensates, reducing their diffusion‑dependent access to chromatin lesions.
3. In progeria, mutant prelamin A forms stable CK2α complexes; we extend this to physiological aging where modest prelamin A elevation still reaches a threshold for LLPS‑mediated kinase trapping.
4. Sequestered kinases fail to phosphorylate downstream substrates (e.g., H2AX, 53BP1, RPA), impairing DDR signaling, fork restart, and R‑loop resolution, which explains the overlap of replication stress and R‑loop accumulation observed in aged tissues.
5. Spermidine rescues progeria by disrupting prelamin A‑CK2α binding; we propose it also interferes with the multivalent interactions that underlie prelamin A LLPS, dissolving condensates and restoring kinase mobility.

Testable Predictions

• Prediction 1: In nuclei from aged human arteries or Lmna L648R/L648R mouse tissues, farnesylated prela

Hypothesis

Baseline senescent cell burden in human tissues exerts a dose‑dependent, biphasic influence on aging phenotypes: low‑to‑moderate levels secrete a SASP gradient that reinforces tissue homeostasis and restrains aberrant proliferation, whereas high burden overwhelms this rheostatic switch, driving chronic inflammation and functional decline.

Mechanistic Rationale

Senescent cells are not uniform offenders; they release a spectrum of soluble factors (IL‑6, IGF‑BPs, TGF‑β, MMPs) whose concentration creates a spatial gradient that neighboring cells interpret via dose‑sensitive receptors (e.g., GP130 for IL‑6 family, TGF‑βR). At low concentrations, these factors activate STAT3 and SMAD pathways that promote extracellular‑matrix remodeling, enhance stem‑cell niche fidelity, and enforce contact‑inhibition‑like signals through up‑regulation of p21^CIP1 in progenitors—effectively acting as chaperones that prevent hyperplasia. When senescent cell density exceeds a tissue‑specific threshold, the same pathways become saturated, leading to persistent NF‑κB activation, ROS amplification, and paracrine senescence spread. This switch is modulated by mechanical cues: stiffened extracellular matrix (common with aging) amplifies YAP/TAZ nuclear translocation, which synergizes with SASP to push cells past the protective tipping point.

Testable Predictions

1. In human cohorts, genetically predicted low‑to‑moderate CDKN2A/p16Ink4a expression (instrumental variable) will associate with better preservation of tissue‑specific functional metrics (e.g., bone mineral density, grip

Hypothesis

Tumor‑derived exosomes enriched in succinate reprogram cancer‑associated fibroblasts (CAFs) to secrete MMP9, thereby remodeling the extracellular matrix, limiting drug penetration and fostering an immunosuppressive niche that drives acquired resistance to HER2‑targeted therapy.

Mechanistic Rationale

• Succinate, an oncometabolite, accumulates in the hypoxic tumor microenvironment and is packaged into exosomes via a mechanism linked to mitochondrial efflux ([1]).
• CAFs express the succinate receptor GPR91; ligand binding stabilizes HIF1α even under normoxia, transcriptionally upregulating MMP9 and TGF‑β1.
• Elevated MMP9 degrades collagen IV and laminin, exposing cryptic ECM fragments that bind CXCR4 on tumor cells, enhancing survival signaling and creating physical barriers that impede trastuzumab penetration.
• Simultaneously, TGF‑β1 skews tumor‑associated macrophages toward an M2 phenotype, suppressing CD8+ T‑cell infiltration—a process documented in metastatic gastric models ([2]).

This proteometabolomic axis explains why genomic profiling alone fails to predict resistance: the driver is a secreted metabolite‑protein circuit, not a mutational event.

Testable Predictions

1. Patients whose baseline plasma exosomes show high succinate‑to‑creatinine ratios will exhibit shorter progression‑free survival on HER2‑targeted regimens.

The protective influence of indole‑3‑propionic acid (IPA) on intestinal mesenchymal cells depends on pregnane X receptor (PXR) activation, yet no study has examined whether PXR expression itself changes with age. We hypothesize that advancing age triggers an epigenetic downregulation of PXR in intestinal fibroblasts and pericytes, rendering these cells unresponsive to IPA despite normal ligand levels. This receptor loss converts a protective signaling circuit into a senescent, pro‑fibrotic state, thereby uncoupling microbiota‑derived metabolites from barrier maintenance.

Mechanistically, age‑associated decline in NAD+ reduces SIRT1 deacetylase activity, leading to hyperacetylation of histone H3 lysine 27 at the PXR promoter. Increased H3K27ac recruits polycomb repressive complex 2, which deposits H3K27me3 and compacts chromatin, silencing Nr1i2 transcription. Consequently, IPA‑bound PXR fails to suppress NF‑κB‑driven IL1β and IL5 secretion, and fibroblast‑derived collagen synthesis proceeds unchecked, reproducing the fibrosis seen in germ‑free or antibiotic‑treated young mice.

The hypothesis is testable and falsifiable. First, quantify Nr1i2 mRNA and PXR protein in isolated intestinal mesenchymal cells from young (3 mo), middle‑aged (12 mo), and old (24 mo) conventionally raised mice; a progressive reduction would support the claim. Second, assess promoter acetylation and methylation states via ChIP‑qPCR for H3K27ac and H3K27me3 at the Nr1i2 locus across the same ages. Third, rescue experiments: treat old mice with the NAD+ precursor nicotinamide riboside or a SIRT1 activ

Aging may persist not as a passive byproduct but as an actively maintained signal that benefits kin by tempering competition for resources. We hypothesize that older individuals release a specific class of extracellular vesicles enriched in miR‑34a and miR‑128 that transiently inhibit germline stem cell proliferation in younger conspecifics. This vesicular signal would be conserved across metazoans, detectable in serum or hemolymph, and its activity would scale with the proportion of post‑reproductive individuals in a population. If true, experimentally reducing vesicle uptake in young animals should lift the reproductive suppression without shortening their lifespan, whereas overexpressing the vesicles in young animals should mimic the reproductive decline seen in older cohorts. Such a mechanism would represent a form of group‑selected senescence that evades the classic cheater problem because the signal’s effect is conditional on the presence of older individuals, making non‑responding mutants vulnerable when older conspecifics are abundant.

To test this, we will use Drosophila melanogaster lines engineered to fluorescently label exosome‑like vesicles and to conditionally knock down the vesicle‑associated Rab‑GTPase Sec15 in the fat body of older flies. First, we will isolate hemolymph from young and old flies and quantify vesicle concentration and miRNA cargo via small‑RNA sequencing. We predict a significant increase in miR‑34a/miR‑128 load in vesicles from old flies compared with young flies (see [3] for baseline vesicle profiling methods). Second, we will generate yo

Hypothesis

Rapamycin's lifespan extension requires functional BAG3‑mediated chaperone‑assisted selective autophagy (CASA) that is activated by reduced cellular mechanical tension downstream of mTORC1 inhibition.

Mechanistic Rationale

• mTORC1 promotes actin polymerization and myosin II activity, raising intracellular tension.
• Inhibition of mTORC1 by rapamycin lowers RhoA‑ROCK signaling, decreasing stress‑fiber formation and overall tension.
• BAG3 contains a mechanical‑sensing domain that binds exposed hydrophobic patches on misfolded cytoskeletal proteins when tension drops, recruiting Hsp70 and LC3 to initiate CASA.
• Thus, rapamycin does not merely upregulate BAG3 expression; it changes the subcellular context that enables BAG3 to act as a mechanostat, shifting clearance from the proteasome to autophagy.
• If BAG3 cannot sense tension (e.g., mutation of its mechanosensing domain), rapamycin‑induced CASA fails, and lifespan extension is lost.

Experimental Plan

• Generate a knock‑in mouse line expressing BAG3 with a point mutation in its mechanosensing domain (e.g., L145P) that abolishes tension‑dependent activation but preserves basal expression.
• Treat wild‑type and mutant cohorts with rapamycin (14 ppm chow) starting at 12 months of age.
• Monitor survival, calculate median and maximal lifespan.
• In parallel, harvest skeletal muscle and heart at 6 months post‑treatment to assess:
  • CASA flux using a tandem fluorescent‑LC3 reporter coupled to a known BAG3 substrate (e.g., mutant desmin).

Hypothesis Chronic elevation of M-CSF in the aged marrow niche triggers a bistable switch: osteoclast precursors expand, secrete further M-CSF, and lock the microenvironment into an osteoclast‑dominant state that physically and chemically excludes erythroid progenitors, even when EPO signaling remains intact.

Mechanistic basis

• Positive feedback loop – EPO acting on CD115+ osteoclast precursors raises M-CSF ~2‑fold within 16 h [1]. Elevated M-CSF promotes osteoclast survival and proliferation, which in turn produces additional M-CSF (autocrine/paracrine). This loop can generate a threshold beyond which osteoclast occupancy exceeds a critical niche volume.
• Niche exclusion – Expanded osteoclasts occupy endosteal niches and secrete proteases (e.g., cathepsin K) that remodel extracellular matrix, reducing fibronectin and VCAM‑1 availability required for erythroid progenitor adhesion and proliferation.
• Iron sequestration – Osteoclast‑derived inflammation elevates hepcidin via IL‑6 signaling, further limiting iron for erythropoiesis [2]. Thus, the niche shift compounds intrinsic erythroid defects through both spatial and metabolic mechanisms.
• Preserved EPOR responsiveness – When osteoclast expansion is genetically blocked (EPOR deletion in myeloid lineage) erythropoiesis is maintained despite niche remodeling [3], indicating erythroid progenitors retain EPO sensitivity if physical competition is relieved.

Testable predictions

1. Threshold detection – In aged mice, M-CSF concentration in marrow interstitial fluid will show a bimodal distribution; values above ~X pg/mL cor

Hypothesis

Protein aggregates in aging cells act as a redox‑regulated filtration system that selectively sequesters oxidatively modified proteins into liquid‑liquid phase separated (LLPS) condensates. Sulfenylation of cysteine residues promotes LLPS by increasing hydrophobicity and promoting multivalent interactions, while persulfidation maintains the condensates in a reversible, chaperone‑rich state. When persulfidation declines with age, sulfenylated clients become trapped, leading to irreversible amyloid‑like aggregation.

Key predictions

1. In young cells, oxidative stress triggers transient sulfenylation of a subset of cytosolic proteins, which drives their partitioning into chaperone‑enriched LLPS droplets that reduce soluble toxic species [1][2].
2. Persulfidation, mediated by hydrogen sulfide (H2S) signaling, coats the same cysteine residues, preventing over‑cross‑linking and allowing rapid disassembly of droplets upon stress relief [5].
3. With aging, mitochondrial H2S production falls, decreasing persulfidation; sulfenylated proteins accumulate in droplets, undergo further oxidation to sulfonylation, and the condensates mature into stable, insoluble aggregates [3][4].
4. Artificially boosting persulfidation (e.g., via NaHS donation) in aged models should dissolve existing aggregates without increasing soluble oxidative damage, whereas inhibiting sulfenylation (e.g., with dimedone) should block aggregate formation even under oxidative stress.

Experimental approach

• Use quantitative cysteine‑oxidation proteomics (OxICAT) to map sulfenylation vs persu

Hypothesis

Circadian regulation of matrix metalloproteinase‑9 (MMP-9) creates daily proteolytic windows that limit the accumulation of non‑enzymatic collagen crosslinks (e.g., glucosepane). Disruption of this clock‑MMP-9 axis accelerates crosslink buildup, increases tissue stiffness, and impairs regeneration.

Mechanistic Rationale

• Core clock output: BMAL1:CLOCK heterodimers drive transcription of Mmp9 via E‑box elements in its promoter, producing a peak of MMP-9 mRNA and protein during the early active phase (circadian time CT6‑CT12)【https://pmc.ncbi.nlm.nih.gov/articles/PMC4241420/】.
• TIMP-1 anti‑phase: Tissue inhibitor of metalloproteinases‑1 (TIMP-1) is repressed by REV-ERBα during the same window, lowering the TIMP-1:MMP-9 ratio and permitting net collagenolysis【https://pubmed.ncbi.nlm.nih.gov/17826643/】.
• Crosslink substrate availability: Long‑lived collagen fibrils are most susceptible to MMP‑9 cleavage when glycation is low; intracellular NAD+ levels, which oscillate with the clock, modulate PARP‑1 activity that can reverse early glycation adducts, reducing the pool of crosslink‑prone lysine residues【https://pmc.ncbi.nlm.nih.gov/articles/PMC7397091/】.
• Desynchronization consequence: Clock gene knockdown or chronic jet‑lag flattens the MMP-9/TIMP-1 rhythm, sustaining a high TIMP-1:MMP-9 bias, thereby decreasing proteolytic clearance of glycated collagen and allowing glucosepane and pentosidine to accumulate【https://scholarscompass.vcu.edu/etd/7485/】.

Predictions and Experimental Design

1. Prediction 1: In wild‑type mice, collagen

Hypothesis\n\nAcarbose extends male lifespan by increasing colonic butyrate production, which activates hepatic FGF21 signaling to inhibit mTORC1 in an androgen‑dependent manner.\n\n## Mechanistic Rationale\n\n- It's known that acarbose blocks α‑glucosidase, delivering undigested starch to the colon where it fuels SCFA‑producing bacteria, raising butyrate levels 3.\n- Butyrate, a histone deacetylase inhibitor, can upregulate Fgf21 expression in hepatocytes via PPARα activation 4.\n- Hepatic FGF21 acts as an endocrine hormone that suppresses mTORC1 signaling, mimicking caloric restriction and rapamycin effects 8.\n- Male mice have higher androgen receptor activity in liver, which synergizes with FGF21 to further dampen mTORC1, whereas females lack this amplification, explaining the modest lifespan gain.\n\n## Testable Predictions\n\n1. Germ‑free or antibiotic‑treated male mice fed acarbose will show no increase in fecal butyrate, no hepatic FGF21 rise, and no lifespan extension.\n2. Oral butyrate supplementation in germ‑free males on acarbose will rescue FGF21 upregulation and lifespan benefit.\n3. Liver‑specific Fgf21 knockout males will abolish acarbose‑mediated mTORC1 inhibition and longevity, despite normal microbiome shifts.\n4. Castration of males will reduce the acarbose‑induced lifespan effect to female‑level levels, while and

Hypothesis

Transient activation of carnitine acetyltransferase (CAT) in senescent muscle cells shifts their metabolism from a persistent fatty‑acid‑oxidation (FAO) state to a balanced acetyl‑CoA buffer, thereby maintaining a reparative SASP and preventing the lipid‑overload that drives chronic inflammation.

Mechanistic rationale

• Senescent myoblasts accumulate long‑chain acyl‑carnitines, indicating excess FAO and insufficient CAT‑mediated acetyl‑group buffering 4.
• Elevated FAO fuels DDR‑mitochondria signaling that raises acetyl‑CoA, cristae remodeling and p16 expression, locking cells in a senescence‑promoting loop 5.
• CAT normally converts acetyl‑CoA + carnitine ↔ acetyl‑carnitine + CoA, providing a sink for excess acetyl groups and linking FAO to the TCA cycle.
• We propose that a decline in CAT activity or Vmax during the acute‑to‑chronic transition creates a metabolic bottleneck: acyl‑carnitines rise, FAO stays high, and the SASP shifts from transient, pro‑regenerative factors to a maladaptive, inflammatory cocktail.

Novel insight

Unlike senolytics that remove the cell, modulating CAT activity preserves the senescent cell’s chaperone‑like signaling while correcting the metabolic inflection point. Early after injury, a boost in CAT activity would limit acyl‑carnitine buildup, sustain a SASP rich in IGF‑1, FGF‑2 and IL‑6 that promotes dedifferentiation of neighboring myoblasts (as seen in salamander limb regeneration) [1](https://www.biotechniques.com/ce

Hypothesis

KDM5B activity sets the immunogenic state of senescent cells by demethylating H3K4me3 at promoters of genes that restrain the cGAS‑STING pathway. When KDM5B is active, H3K4me3 levels at these loci stay low, keeping TREX1, DNase2 and other STING antagonists expressed, which dampens interferon signaling and yields a SASP rich in tissue‑repair factors. Loss of KDM5B leads to H3K4me3 accumulation at the same promoters, silencing antagonists, unleashing STING‑TBK1‑IRF3 signaling and shifting the SASP toward a pro‑inflammatory, immunogenic profile that recruits immune clearance.

Mechanistic Reasoning

1. Chromatin gatekeeping – KDM5B removes activating H3K4me3 marks. Recent work shows histone methylation balance gates whether cells broadcast danger signals or stay immunologically silent (KDM5 histone demethylases suppress cGAS‑STING).
2. Target loci – Bioinformatic analysis of senescent fibroblast ATAC‑seq and ChIP‑seq datasets predicts KDM5B binding near TREX1, DNase2 and IFI16 promoters, all negative regulators of cytosolic DNA sensing.
3. Signal output – Low antagonist expression permits cGAS activation by self‑DNA, driving TBK1‑IRF3‑dependent IFN‑β production and NF‑κB‑dependent SASP components (IL‑6, IL‑8). High antagonist expression blocks this cascade, favoring a SASP enriched in TGF‑β, VEGF and IGF‑1 that supports tissue remodeling without strong immune recruitment.
4. Negotiator switch – Thus senescent cells can toggle between a “repair‑negotiator” mode (high KDM5B activity) and an “al

Hypothesis

Chronic autophagy in aging intestinal stem cells (ISCs) does not merely recycle intracellular nutrients; it actively degrades extracellular Wnt ligands and niche‑stabilizing proteins, thereby collapsing the Wnt gradient that maintains stemness. This positions autophagy as a rheostat that, when persistently active, converts a protective recycling program into a niche‑destructive force.

Mechanistic Rationale

• Autophagy is induced by AMPK activation and mTORC1 inhibition during nutrient stress, directing LC3‑positive phagophores to specific cargo via receptors such as p62 and NBR1 {https://pmc.ncbi.nlm.nih.gov/articles/PMC9652773/} {https://pmc.ncbi.nlm.nih.gov/articles/PMC4871809/)}.
• In senescent cells, constitutive mTORC1 signaling becomes unresponsive to starvation, sustaining a futile loop where autophagy‑derived amino acids fuel anabolism despite external depletion {https://doi.org/10.1083/jcb.201610113}.
• We propose that, under prolonged siege conditions, autophagosomes also sequester secreted Wnt agonists (e.g., Wnt3a, R‑spondin1) and extracellular matrix components that sequester Wnt inhibitors (e.g., DKK1, sFRPs). Degradation of these niche‑supporting factors reduces extracellular Wnt availability, flattening the crypt gradient.
• Loss of Wnt signaling pushes ISCs toward differentiation or apoptosis, while residual mTORC1 activity—fed by autophagy‑generated metabolites—drives maladaptive proliferative

Hypothesis

Rapamycin’s inhibition of mTORC1 triggers autophagy and downstream FOXO/ATF4 signaling, which are classically viewed as a hormetic stress response that mimics caloric restriction Rapamycin extends lifespan primarily by slowing damage accumulation Both rapamycin and CR inhibit mTORC1, but their effects are age-, tissue-, and time-of-day dependent. It's plausible that the epigenetic mark recruits co‑activators that maintain FOXO transcription. Yet persistent benefits after drug withdrawal suggest that the pathway may also write a lasting epigenetic program that sustains a youthful transcriptional state. We hypothesize that rapamycin induces a stable chromatin configuration at loci governing proteostasis and mitochondrial function, converting a transient stress signal into a durable repair program. This epigenetic memory would be detectable as sustained histone acetylation and reduced DNA methylation at FOXO‑target promoters long after the drug is cleared, and it would correlate with continued improvement in tissue function even when mTOR activity returns to baseline.

Predictions

• Histone H3K27ac levels at FOXO‑dependent enhancers remain elevated for at least four weeks after rapamycin cessation in mouse liver and heart.
• DNA methylation at CpG islands of FOXO3 and SIRT1 promoters shows a significant decrease that persists post‑treatment, whereas global methylation patterns revert to pretreatment levels.
• Animals treated with rapamycin for a short w

{"title":"TET2/3-mediated 5hmC erosion at metabolic‑gene enhancers drives age‑associated T cell dysfunction independent of global DNA methylation","body":"# Hypothesis\n\nAge‑related decline in T cell function stems not from random DNA methylation drift but from progressive, locus‑specific loss of 5‑hydroxymethylcytosine (5hmC) at enhancers governing metabolic reprogramming. This erosion is catalyzed by TET2 and TET3 activity, which, paradoxically, reduces 5hmC at these sites over time, leading to enhancer hypo‑hydroxymethylation, weakened transcription of glycolysis and oxidative‑phosphorylation genes, and a shift toward a senescent, low‑energy phenotype.\n\n## Rationale\n\n- TET2/3 oxidize 5mC to 5hmC at enhancers and promoters during thymic development and effector differentiation, stabilizing genes such as Zbtb7b [1].\n- In CAR‑T cells, TET2 loss limits exhaustion and boosts persistence by fostering a memory phenotype [2], indicating that TET2 activity can modulate differentiation trajectories.\n- Genome‑wide 5hmC mapping shows enrichment in gene bodies and enhancers of active T cell genes, with tissue‑specific patterns stable early in tumorigenesis [3] and a joint‑snhmC‑seq protocol enabling simultaneous 5hmC, 5mC, and transcriptome profiling [4].\n- Gene‑body 5hmC accumulates with age in multiple mouse tissues and correlates w