Targeted senolytic prodrug is well tolerated and results in amelioration of frailty, muscle regeneration and cognitive functions in geriatric mice


 Mammalian aging coincides with an accumulation of senescent cells. Depletion of senescent cells shows promise to treat age-related diseases. However, many of the senolytic drugs used to deplete senescent cells cause profound toxic side effects. Here we tested a prodrug form of a pan-cytotoxic drug and demonstrated dramatically reduced side effects in geriatric mice while gaining senolytic selectivity. Specifically, we employ a strategy that takes advantage of the greatly enriched expression of the lysosomal hydrolase β-Galactosidase in senescent cells as a tool to selectively convert the prodrug into the active parent drug. In young, old and geriatric mice this prodrug was well tolerated and proved capable of reducing the burden of senescence in multiple tissues without evident toxicity. Importantly, chronic systemic treatment of geriatric mice resulted in reduced frailty index, improved muscle tissue and stem cell functions, improved cognitive functions and overall improved survival. This work supports future senolytic prodrug designs, based on other metabolic functions characteristic of senescent cells, on different parent drugs and aimed at clinical development. This new class of senolytic prodrugs promises capabilities of targeting selectively and safely specific senescent cell types in multiple tissues to treat age-related diseases in geriatric people.


Introduction
Frailty is one of the most relevant clinical syndromes in geriatric patients 1,2 , which entangles older adults in a vicious cycle of energy dysregulation 3 leading to loss of weight, strength and balance, exhaustion, low exercise capacity, and slow, unstable walking. The outcome is an increased state of vulnerability, which diminishes the ability of the body to respond to stress, thus reducing physiological reserves and impairing the ability of the body to maintain homeostasis 4 . In frail individuals, an apparently small insult (e.g., introduction of a new medication; minor infection; or minor surgery) can result in a dramatic and disproportionate change in health state: from independent to dependent; mobile to immobile; postural stability to falling; lucid to delirious [5][6][7] . Frailty is also linked with cognitive decline, dementia and Alzheimer's disease (AD) 4,5 . Although Frailty has been recognized by the World Health Organization 1 and the U.S. National Academy of Sciences 8 as a major health concern among the elderly, an effective therapeutic solution is still not available.
Frailty is connected to cellular aging, which in turn is connected to cellular senescence 9 . Senescent cells are permanently withdrawn from the cell cycle and generally develop a persistent pro-inflammatory phenotype called the senescence-associated secretory phenotype (SASP) which is comprised of proinflammatory cytokines and chemokines 10 . Accumulation of senescent cells has been observed at sites of many agerelated injuries 11 . Studies confirmed that injecting senescent cells into young mice increased frailty and mortality 12 . Inducing apoptosis of senescent cells in vivo delays deterioration of the vasculature, heart, and kidney, and it replenishes the hematopoietic stem cell pool restoring tissue homeostasis after chemotoxicity and aging [13][14][15][16] . The efficacy of this treatment strongly depends on the subtype of the senescent cells targeted, being most pronounced against senescent cells with a high inflammatory status 17 .
Selective killing of senescent cells with therapeutics (i.e., senolytics) have gained attention as a new therapeutic approach for age-related diseases. Targeting of prosurvival Senescent Cell Anti-apoptotic Pathways (SCAPs) 18 has emerged as the primary strategy for senescent cell killing. Until now, the reported classes of senolytics include kinase inhibitors, Bcl-2 family inhibitors, natural products, a p53 binding inhibitor, heat shock protein 90 (HSP90) inhibitors, UBX0101, a histone deacetylase (HDAC) inhibitor, cardiac glycosides and bromodomain and extra-terminal domain (BET) family protein inhibitors 19 . The combined administration of dasatinib and quercetin (targeting BCL-2/BCL-XL, PI3K/AKT, p53/p21/serpine and the dependence receptors/tyrosine kinase SCAP) by clearing senescent cells in the tissues has been shown to improve the pathologies of diverse age-related diseases, including cardiac aging 19 , atherosclerosis 20 , osteoporosis 21 , pulmonary fibrosis 22 , hepatic steatosis 23 , and Alzheimer's disease 24 . However, the general lack of specificity towards senescent cells 25,26 often leads to hematological toxicities including neutropenia 27 and thrombocytopenia 28 . Moreover, mTOR inhibitors have found their way into senolytic application since an inhibitory action on SCAPS has been described 29 . Again, their translational value in vivo is limited due to their chronic toxicity, which leads to hyperglycemia, hyperlipidemia, insulin resistance and increased incidence of new-onset type 2 diabetes 30 . These specificity issues of current senolytics limits their application to local delivery and/or to acute treatments as opposed to chronic systemic administration. The identification of agents that selectively kill senescent cells while sparing other cell populations represents a scientific challenge. Current senolytic drugs target molecular pathways shared between senescent and proliferating cells, thus achieving cell killing but not specificity. As a matter of fact, many known senolytic agents were initially developed as cytotoxic anti-cancer agents and subsequently repurposed for 'selective' removal of senescent cell populations. Because proliferating cells are frequently more sensitive to the cytotoxic or cytostatic effect of anti-tumor agents, dose-limiting toxicity, especially in rapidly replicating hematopoietic, skin or gut cells, is a frequently observed side-effect, which strongly limits the clinical utility of these anti-senescence therapies 27 . Pulsatile administration of senolytic drugs has been proposed as a mechanism to minimize exposure of non-senescent cells and therefore off-target effects 31 .
However, this approach likely limits therapeutic efficacy due to restricted dosing regimens, particularly in elderly individuals affected by geriatric syndrome. Typically, this patient population is frail, affected by comorbidities and do not tolerate the side effects sustained by younger individuals. Currently, there remains an unmet need to identify senolytics that can be targeted to senescent cells safely, selectively and systemically. Senescent cells are characterized by a notable change in biological properties such as an increase in the levels of mitochondria, reactive oxygen species, lysosomal content, and upregulation of many lysosomal proteins, including the lysosomal enzyme senescence-associated β-galactosidase (SA-βGal) 32,33 . Increased activity of SA-βgal is detected not only in cells undergoing senescence in vitro, but also in the tissues from aged animals and humans [34][35][36] . Recently, a promising strategy has been proposed based on galactose-derivative prodrugs. These prodrugs are selectively activated in senescent cells upon conversion into the parent active drug by the hydrolase activity of SA-βGal. In particular, specific senotoxic compounds such as duocarmycin, gemcitabine and navitoclax have been modified into galacto-derivative prodrugs showing increased selectivity in targeting senescence cells and efficacy in treating cancer and aged mouse models [37][38][39] .
Here, we report a novel prodrug design to target senescent cells, allowing systemic removal of senescent cells in geriatric mice without noticeable side effects. We took advantage of the senescence-specific activity of SA-βGal in the design of a non-toxic senolytic prodrug derivative of the compound 5-Fluorouridine, a metabolic precursor of the clinically approved anti-cancer medication 5-Fluorouracil. We first tested the specificity of this prodrug on senescent cells in vitro. We then confirmed safety and efficacy of the prodrug in young (5 month-old), aged (22 month old) and in geriatric (30 month old) mice.
Importantly, we showed that geriatric mice that received the prodrug treatment for four weeks altogether improved significantly: 1) frailty profile; 2) skeletal muscle function; 3) muscle stem cell function; 4) cognitive function; and 5) survival. Prodrug senolytic pharmacological treatments are a promising strategy to overcome the current limitations in chronic and systemic treatments of older individuals affected with neurodegenerative diseases, age-related diseases and geriatric syndromes.

Generation of 5FURGal Prodrug.
Design of the senolytic prodrug employed in this study begins with a small molecule having intrinsic senotoxicity, and then chemically conjugates at an appropriate site within the molecule the hexose sugar galactose to generate a β-D-galactopyranosyl adduct such that the cytotoxicity of the parent molecule is masked. The inactive prodrug is therefore transformed into its cytotoxic parent only within senescent cells after hydrolytic cleavage of the β-D-galactosyl promoiety by the abundantly expressed SA-βGal enzyme (FIG 1a). A small library of cytotoxic chemotherapeutic compounds amenable for chemical modification as described was assembled. In vitro screening was performed to verify the ability of these small molecules to induce apoptosis in human senescent fibroblasts. 5-Fluorouridine (5FU) is a well-characterized chemotherapeutic agent, which exerts its cytotoxic action on cancer cells by interfering with DNA replication through inhibition of thymidylate synthase. Interestingly, 5FU also showed the ability to induce apoptosis in senescent cells by interfering with RNA synthesis 40 . This compound is known for its high level of toxicity in vivo and was chosen to test the capacity of the prodrug derivation to prevent its side effects while retaining senolytic efficacy. Following our prodrug design concept, we modified 5FU into the inactive prodrug 5-fluorouridine-5′-O-β-Dgalactopyranoside (5FURGal) and proceeded to validate its senolytic activity and safety in vitro and in vivo (Fig. 1b).

5FURGal prodrug selectively kills senescent cells in vitro.
We first tested the specificity of the 5FURGal prodrug. In particular, we evaluated whether the pro-apoptotic effects of the parent drug 5FU against proliferating or quiescent human fibroblasts were blunted for prodrug 5FURGal, while maintaining senotoxic efficacy of the latter against senescent fibroblast cells.
As anticipated, while the cytotoxic potency of 5FU was greatest against proliferating human IMR90 fibroblasts (IC50 ~ 0.0015 mM), substantial toxicity of the parent drug was also observed against quiescent or senescent fibroblast cells (IC50 = 0.07 and 0.039 mM respectively). By contrast, we found that 5FURGal showed minimal toxicity against either proliferating fibroblasts (IC50 ~ 4.4 mM) or quiescent fibroblasts (IC50 > 10 mM) but retained pro-apoptotic activity toward senescent IMR90 cells (IC50 ~ 0.30 mM) as intended (Fig 1c-e). To confirm the hypothesized mechanism of action that 5FURGal is specifically hydrolyzed by SA-βGal, we employed human fibroblasts where the gene GLB1 encoding β-Gal has been knocked out. As predicted, we found that β-Gal deficient senescent cells were significantly less sensitive to 5FURGal (IC50 > 10 mM) than β-Gal proficient senescent cells (IC50 ~ 0.30 mM), suggesting loss of their ability to convert 5FURGal into parent drug 5FU ( Fig. f-g). Finally, to explore non-specific effects of 5FURGal on the cell cycle, we cultured proliferative or senescent fibroblasts, with or without 5FURGal and in the presence of EdU, a thymidine analog that is incorporated during the S Phase of the cell cycle. We found that 5FURGal had no impact on cell cycle phases in either cell population (Supp. Fig. 1). Taken together, these results show that 5FURGal is efficiently converted into 5FU in senescent cells but not in proliferating or quiescent cells, suggesting that prodrug 5FURGal might be a suitable candidate senolytic molecule with high selectivity for senescent cells and minimal toxicity for non-senescent cells.

5FURGal is well tolerated in vivo.
To evaluate whether the selectivity of 5FURGal can prevent the toxicity of the parent drug 5FU, we tested the effects of both compounds in vivo in four month-old C57/b6 mice.
We found that 5FURGal did not replicate the severe acute toxicity of 5FU observed after IP administration. In particular, dosing mice with (100 mg/kg) 5FU resulted in: a) loss of body weight (Fig. 1h); b) loss of spleen mass (Fig. 1i), loss of bone marrow cell count (Fig. 1l), and c) thrombocytopenia, neutropenia, and lymphocytopenia ( Figure 1m). Conversely, mice treated with an equimolar dose of 5FURGal (160 mg/Kg), similar to vehicle-treated control mice group, showed no deleterious impact on any of the parameters analyzed, suggesting that 5FURGal is not acutely toxic in the tissues tested. Moreover, pharmacokinetic studies demonstrate that mice dosed intravenously with 5FURGal show low levels of systemic conversion of prodrug to the parent drug 5FU (~2% on a dose-adjusted basis) (Supp Fig.   2). Taken together, these results indicate that 5FURGal could be safely administered systemically with minimal toxicity and high tolerability in young mice.
A luciferin prodrug is employed to assess senescence-associated bioluminescence in aged mouse models.
To test the senolytic efficacy of 5FURGal, we investigated its ability to target senescent cells in vivo. First, we developed a novel preclinical murine model to assess senescence levels non-invasively using a prodrug strategy similar to 5FURGal. Young (4 months) and aged (22 months) transgenic mice constitutively expressing the bioluminescent enzyme Luciferase (FVB-Tg(CAG-luc,-GFP)L2G85Chco/J or FVB-LucGFP) were used to develop a method of senescence-associated bioluminescence imaging (SA-BLI). To achieve this, we employed LUCGal, a β-D-galactosyl conjugated prodrug version of luciferin (the luciferase substrate). Similar to 5FURGal, LUCGal is hydrolyzed selectively in cells where expression and activity of SA-βGal are highly enriched (Fig. 2a). Bioimaging of these mice showed that aged mice have ~6 fold higher levels of bioluminescence compared to young mice, suggesting increased levels of SA-βGal + cells associated with age ( Fig. 2b and   c). As a positive control, we treated a cohort of young mice with the chemotherapeutic agent Doxorubicin ('chemo mouse' model), reported to induce cellular senescence in vivo and capable of promoting an accelerated aging phenotype in cancer patients 41,42 . We found that the levels of SA-BLI in chemo mice were similar to naturally aged mice ( Fig. 2b and c). Next, from the same mice previously employed for the SA-BLI studies we isolated liver tissue to confirm levels of senescence histologically. We performed colorimetric staining for SA-βGal in sections of liver and we found that, similar to the SA-BLI results, young mice exhibited only baseline levels of SA-βGal. Conversely, both chemo mice and naturally aged mice showed ~800-fold higher levels of SA-βGal compared to baseline, indicating accumulation of senescent cells ( Fig. 2d and e). These results were consistent with gene expression levels of Cdkn2A (p16 INK4A ), another reported marker of cellular senescence. We performed qPCR in liver samples isolated from the same mice, and we found that p16 INK4A was upregulated ~16 fold in naturally aged mice and young chemo mice compared to untreated young mice ( Fig. 2f). Taken together, these results suggest that a prodrug derivative of luciferin can be effectively employed to achieve selective SA-BLI in aged mice as a preclinical model to study cellular senescence.

5FURGal shows senolytic and functional efficacy in aged mice.
To test efficacy of the prodrug 5FURGal for senolytic and functional activity in vivo, we evaluated the overall physical performance for aged FVB-LucGFP mice and correlated this with SA-BLI levels. To this end, we measured grip strength, endurance, balance and spatial learning memory and combined the results into a performance score. We found that this performance score inversely correlates with SA-BLI levels in aged FVB-LucGFP mice, suggesting that in aged mice higher levels of senescence are associated with lower physiological performance (Fig. 3a). Next, we treated aged FVB-LucGFP mice via intraperitoneal (IP) injections with two doses (separated by two days) for two weeks of: a) 5FURGal at 140 mg/Kg; or b) saline. Mice treated with 5FURGal showed a reduction of SA-BLI when measured 3 days following the last injection (Figure 1b and c). Next, the day after SA-BLI was measured we tested the effect of 5FURGal treatment on physiological functions.
Compared to aged mice treated with saline, aged mice treated with 5FURGal showed improved grip strength, running time and learning spatial cognitive functions ( Fig. 3 d-f). To confirm prior SA-BLI results, we histologically assessed SA-βGal and we found lower levels in livers of aged mice treated with 5FURGal ( Fig. 3h, i). Similarly, p16 INK4A gene expression levels were reduced in kidney and liver tissues (Figure 1g and l), further suggesting that 5FURGal treatment results in clearance of senescent cells in vivo. To confirm these results in a different model of in vivo senescence, we employed the chemo mouse model. Chemo mice that received a treatment with 5FURGal 4 and 6 days after the last Doxorubicin injection presented reduced levels of SA-βGal staining in the liver compared to chemo mice treated with saline (Supp. Fig. 3a,b). Taken together, these results suggest that 5FURGal is capable of reducing senescent cell levels in vivo in multiple tissues and following different senescence induction stimuli. Finally, to test the beneficial senolytic effects of 5FURGal in a model different from natural aged mice, chemo mice were employed as a model of accelerated aging. When treated with saline, chemo mice showed significant weight loss 8 days following Doxorubicin dosing. However, 5FURGal treatment increased the tolerability for Doxorubin treatment with a significant attenuation of weight loss (Supp. Fig. 2c). Taken together, these results suggest that the beneficial senolytic effects of 5FURGal in aged mice models lead to recovery of youthful healthy physiological traits.

5FURGal reverts physiological traits of Frailty in geriatric mice.
Next, we aimed to test whether 5FURGal was able to reduce the burden of Frailty in geriatric mice. With this goal in mind, we chose a model of aged-related Frailty Syndrome which includes loss of certain skeletal muscle functions. In particular, we employed a cohort of geriatric (30 month-old) Pax7 CreER /flox-YFP mice, that express the yellow fluorescent protein (YFP) specifically in muscle stem cells upon induction with Tamoxifen. We employed this cohort of geriatric mice for extensive longitudinal studies on physiological and cognitive functions. To comprehensively study the effects of 5FURGal on geriatric mice, the experiment was designed in three phases: 1) Frailty Assessment; 2) Cognitive Assessment; 3) Skeletal Muscle Function Assessment (Fig. 4a). The mice were randomly divided into two groups, one to be treated with saline (vehicle control) and one with 5FURGal (dose 140 mg/Kg), twice per week for four weeks in total, with a week of rest in between (Fig. 4a). In Phase One, we assessed the mice functionally before and after the treatment. A characteristic of Frailty is exhaustion and overall weakness, which results in decreased mobility that is considered to be one of the best predictors of age-related loss of functions 43 .
To test this, mice were housed individually in cages with access to computer-controlled running wheels during the entire duration of Phase One. We found that geriatric mice treated with 5FURGal significantly improved their spontaneous activity, quantified as total distance run over time. Conversely, geriatric mice treated with saline showed the opposite trend ( Fig.   4b and Supp Fig. 4a). These results suggest that 5FURGal treatment improves activity levels in geriatric mice.
To generate a Frailty index, we measured 10 different parameters in young mice to set the youthful baseline level (defined as 100% score, where each parameter was weighted for 10% of the score). In particular, we measured: 1) body weight; 2) motor coordination; 3) endurance; 4) dystonia; 5) ataxia; 6) grip strength; 7) spontaneous activity; 8) cognitive functions; 9) gait; and 10) coat condition. Frailty score was assessed in the geriatric mice cohort before treatment, and we found that they scored an average ~30% (geriatric level) compared to the youthful basal level. However, when we repeated the measurements following treatments, we found that geriatric mice receiving 5FURGal improved their Frailty Score, averaging ~60%, while saline-treated geriatric mice did not show improved scores ( Fig. 4c). Moreover, despite the small cohort analyzed, geriatric mice treated with 5FURGal showed a trend of ~40% higher survival and lower mortality compared to mice treated with saline ( Fig. 4d). Taken together these results suggest that systemic senolytic treatment with prodrug 5FURGal results in amelioration of the Frailty profile and decreased mortality in geriatric mice.

Prodrug treatment improves cognitive functions in geriatric mice.
Loss of memory and loss of spatial orientation are two aspects characteristic of agerelated dementia and Alzheimer's disease (AD) which are often comorbid with Frailty Syndrome 44,45 . In Phase Two, we investigated 5FURGal properties in reversing cognitive decline. To measure spatial learning, we employed the Y Maze test. Mice were tested before and after treatment. Treated mice with 5FURGal improved their spontaneous alternation performance significantly compared to mice treated with saline, suggesting improved spatial learning functions ( Fig. 5a and Supp Fig. 5a-b). We then performed the Morris Water Maze test to measure learning memory and memory retrieval. Geriatric mice treated with 5FURGal showed shorter latencies to escape onto the hidden platform and spend more time in the target quadrant in the probe test performing more similarly to young mice, suggesting improved memory functions (Fig. 5b, c). The loss of cognitive functions associated with age is linked to morphological changes in the hippocampus, where microglial cells' activation increases 46,47 . Brains of mice were fixed, sectioned and immune-stained with antibodies against IBA-1 (marker of microglia cells) and CD68 (marker of activated microglia). We found that increased levels of microglia and activated microglia observed in the hippocampal brain regions of geriatric mice treated with saline were significantly reduced after 5FURGal treatment and more similar to brains of young mice (Fig.5 d-f). Taken together, these results show that systemic senolytic treatment with 5FURGal improved cognitive functions and reduced neuroinflammation in geriatric mice.

Prodrug treatment restores skeletal muscle regenerative capacity and stem cell functions in geriatric mice.
Frailty Syndrome is associated with loss of muscle strength and sarcopenia, which is associated with loss of muscle stem cell (MuSC) function 48,49 . In Phase Three, we measured skeletal muscle functions, including MuSC functions and muscle regenerative capacity after inducing an injury performing intramuscular injections of Cardiotoxin (CTX) into tibialis anterior (TA) muscles. First, to trace MuSCs over time, we injected young and geriatric mice with Tamoxifen to induce the recombination and expression of the YFP protein driven by the Pax7CreEr locus, which is selectively expressed by MuSCs. Then, to assess muscle contractile properties, we performed ex vivo electrophysiological biomechanics measurements of dissected TA muscles 10 days after CTX injury. We found that TA muscles from geriatric mice treated with saline generated ~60% less contractile tetanic force compared to muscles from young mice treated with saline. Conversely, geriatric mice treated with 5FURGal significantly improved tetanic force generation compared to saline treatments ( Fig. 6a). We then asked if the improved functionality of skeletal muscle was associated with improved regenerative capacity. To that end, we performed immunohistochemistry on sections of the left TA muscles analyzed previously for contractile properties. Morphometry analysis showed that the cross-sectional area of YFP + centrally nucleated myofibers, a hallmark of regenerated myofibers, was significantly increased in geriatric mice treated with 5FURGal compared to geriatric mice treated with saline (Fig. 6b, c).
Next, we wanted to explore whether the improved muscle regenerative capacity was associated with improved MuSCs functions. Following muscle function assessment, YFP + MuSCs were isolated from the right TA muscles by fluorescence-activated cell sorting (FACS). With age, MuSCs decline in number and in their proliferative potential 50 . We found that MuSCs isolated from geriatric mice treated with 5FURGal were higher in number compared to treatment with saline ( Fig. 6d). Moreover, when these MuSCs were cultured in vitro, they exhibited enhanced proliferative capacity analyzed in a single cell myoclonal formation assay and by quantification of EdU + YFP + MuSCs, after culturing them with EDU ( Fig. 6e-f). Taken together, these results indicate that systemic senolytic treatment of 5FURGal in geriatric mice improves skeletal muscle functions, MuSC functions and regenerative capacity.

Chronic systemic treatment with prodrug is well tolerated in geriatric mice.
Finally, we determined the tolerability profile of 5FURGal dosed chronically in geriatric mice. We thus performed toxicology studies in a separate cohort of geriatric mice (C57/b6 mice 28-30 months old). Geriatric mice received a total of five tail vein injections with 130 mg/Kg of 5FURGal or saline over a three-week period. Similar to the Pax7CreEr/YFP mice cohort, no significant difference in body weight was detected in the two groups (Supp. Fig. 6a). Additionally, 5FURGal treatment did not negatively affect mice survival over the course of the three-week treatment (Supp. Fig. 6b). Spleen was dissected and analyzed at the end of the experiment (day 21). Although no statistical difference was measured between 5FURGal or saline treatments, the former showed a trend toward smaller spleen sizes, similar to young mice. Conversely, geriatric mice treated with saline showed aged-related enlarged spleen sizes (Supp. Fig. 6b). The hematology profile of blood isolated from both groups at day 21 did not show significant differences across a multitude of hematopoietic cell types (Supp. Fig. 6c). Taken together, these results suggest that, regardless of the elevated toxicity of the parent drug 5FU, the prodrug 5FURGal is well tolerated in geriatric mice and can be safely administered systemically and chronically without evident side effects.

Discussion
Frail old individuals often present with one or more comorbidities, including myocardial infarction, peripheral vascular disease, stroke, cancer, diabetes mellitus, arthritis, and chronic lung disease 51 . Due to their altered health status they cannot tolerate the side effects previously described for current senolytic therapies. For this reason, most of the current senolytic drugs are under development for single aged-related conditions. We explored a prodrug-based approach to increase the selectivity and the tolerability of a senolytic agent, with the goal of developing strategies for safe systemic therapeutics. In particular, we took advantage of the well-known SA-β-Gal, an hydrolase enzyme enriched in the lysosomal compartment of senescent cells and employed as a marker for cellular senescence. To this end, we employed a 5′-O-β-D-galactopyranosyl modification of the toxic nucleoside 5-fluorouridine 5FUR to generate the selective senolytic prodrug 5FURGal, and have investigated this molecule in various rodent models of age-related pathologies. We found that the prodrug 5FURGal was able to kill human senescent cells but spared proliferative or quiescent cells in vitro. When tested in vivo, young, old and geriatric mice tolerated the prodrug well with no evident toxic effects. Conversely, the parent drug 5FU, a toxic chemotherapeutic agent, is poorly tolerated in young mice and not tolerated in aged mice. As a senolytic, prodrug 5FURGal reduced the level of senescence analyzed through multiple methods, including a non-invasive in vivo bioluminescent assessment based on the same galacto-derivative prodrug strategy. Finally, efficacy studies of 5FURGal in geriatric mice demonstrated that the prodrug reduces the frailty profile while improving: 1) muscle tissue and stem cell functions; 2) cognitive functions; 3) and overall mouse survival. Taken together, our results show that using a sugar derivative prodrug-based strategy we can turn a pan-cytotoxic drug into a safe and effective senolytic. This strategy supports a new generation of senolytic drugs to overcome current limitations in treating elderly individuals with geriatric syndromes.
Aging is often characterized by skeletal muscle catabolism leading to sarcopenia and frailty. Sarcopenia is a syndrome characterized by progressive loss of skeletal muscle mass and function related to aging, disease, starvation or disuse, and typically precedes frailty.
Frailty is an age-related syndrome comprising states such as exhaustion, weakness, and slowness, and more broadly, cognition. Recent studies indicate a significant relationship between age-associated declines in both physical and cognitive function 52 . Importantly, frailty can be considered to be a state of diminished resolution of homeostasis following a stressor event that increases the risk of adverse health outcomes, including fall-related injuries and even delirium. Navitoclax, a Bcl-2 class inhibitor developed for cancer therapy, has been shown to increase the parent drug senolytic specificity and reduce the platelet toxicity 39 ; 3) a galacto-derivative prodrug of gemcitabine, another chemotherapeutic agent, has been reported to reduce inflammation and restore physical function in 20 month old mice 38 . These reports are consistent with our findings that a galacto-derivative prodrug is a viable strategy to turn a pan-cytotoxic small molecule into a selective agent capable of killing senescence cells. With this goal in mind and focusing on the need of minimizing the side effects and toxicity of drugs in geriatric patients, we tested the prodrug 5FURGal, derivative of the cytotoxic agent 5FU. Our results show that this is possible. Importantly, not only young or older mice but also geriatric mice tolerated 5FURGal without noticeable side effects. That allowed us to treat geriatric mice, typically frail and very sensitive to compound side effects, dosing them with chronic systemic regimens that resulted in improved physiological functions at multiple levels, including cognitive and muscular functions, and overall improved survival.

Conclusions
We believe that observations and strategies developed during this study will serve as a valuable precursor to generating novel and more potent prodrugs for systemic senolytic applications in the clinic. The galactose-derivative prodrug 5FURGal evaluated in this study as a tool compound has an improved safety profile due to its superior specificity against senescent cells. However, beyond SA-β-Gal, we have identified other metabolic functions that are preferentially enriched in senescent cells. In our drug development endeavours, we are currently employing these activities to produce a new generation of prodrugs designed for clinical use. Importantly, these novel prodrugs are based on parent drugs, different from 5FU, that we found to have intrinsic selective senolytic activity and superior potency prior to the prodrug derivation. By implementing these compounds through additional medicinal chemistry optimization we aim to eliminate specific senescent cell types accumulated during aging and pathogenic processes without affecting other cell populations, including proliferative or non-senescent quiescent cells (such as many stem cells located in their niche). This approach could allow reduction of the chronic inflammatory process associated with senescent cells, which in future clinical developments will be tested for the efficacy in reducing the burden of physical pathology (such as fibrosis or cancer), cognitive decline (such as dementia or Alzheimer's disease) and the degeneration associated with Frailty Syndrome and other age-related diseases.      Muscles from young or geriatric mice were harvested after taking force measurements (scale bar = 500 µm). c. Quantification of the cross-sectional area of centrally-nucleated YFP+ fibers in regenerating TA muscles that received CTS injections (n = 4). d.

Figure Legends
Quantification using fluorescent activated cell sorting (FACS) of MuSCs YFP+ isolated from injured TA muscles that received CTW injections (biological replicates n = 4). e.
Myoclonal assay of MuSCs YFP+ isolated from injured TA muscles that received CTW injections. Single MuSCs were isolated using FACS in a 96 well plate and quantified for their capacity to generate colonies (biological replicates n = 4). f. Cell cycle assessment.

Pharmacokinetic studies
Concentrations of 5FU and 5FURGal were measured in plasma after IV dosing to male CD1 mice (N=3/dose group) at Pharmaron, Inc., Beijing, China. Compounds were formulated in 15% EtOH/15% PEG400/70% saline and dosed at 5 mg/kg and 40 mg/kg respectively. Blood was sampled at 8 time points over 24 hrs, quenched with an equal volume of 25 mM citrate buffer, and processed for plasma. Concentrations of compounds in plasma were determined using a specific LC-MS/MS method using a Shimazdu DGU-20A5R HPLC, equipped with a 50 x 3 mm Phenomenex Synergi 2.5 mm Polar-RP column and an API 5500 LC-MS/MS detector. Compounds were eluted using a mobile phase gradient comprising water/MeOH/0.1% formic acid, and concentrations determined relative to calibration standards. Pharmacokinetics of 5FU and 5FURGal were calculated using WinNonlin (Phoenix, version 6.1).

Voluntary running.
Each subject was singly housed in cage with computer-controlled running wheels. All animals had access ad libitum to food, water and wheel running. Prior to the experimental period, subjects were given 3 days to acclimate to the cages and their surroundings. At day 3, the experimental group had internal ablations performed on both legs as described in 'Ablation Model' section in 'Methods'. Mice were then returned to their cages with computer-controlled running wheels and allowed to run freely. Non-exercised controls were singly housed in cages with locked wheels. Cycles of 12-h day per night light were used in the rooms housing the mice and their cage bedding was changed once a week.
Each cage wheel was attached to a counter motor that was then attached to data acquisition software that recorded the number of rotations that occurred during each 15 min interval for 2 weeks.

Treadmill running.
Before each mouse began its exercise programme, it was acclimated to the Exer 3/6 treadmill (Columbus Instruments) for 3 days with a 10 min running period at 10mmin each day. At the start of each training regimen, the mice slowly acclimated to the speed of the treadmill by beginning at 7 m/min and increasing the speed at 0.5 m/min until a speed of 12 m/min was reached. On the fourth day, mice were tested for endurance, running at a speed of 13 m/minute, with subsequent increments of 1 m/minute every minute. A low voltage was applied to a grate at the end of the treadmill to encourage the mice to continue running for the duration of the protocol. Mice were removed from the treadmill at the point of exhaustion, which was defined as staying 5 seconds on the electric grid 61 .

Grip test
Mice were put on a BioSeb grip tester and their grip force measured for all four limbs. Mice were measured three-five times and the average was used to calculate grip strength.
Values were normalized to mouse weight.

Rotarod performance test
Rotarod rotating cylinder (rod) (IITC Life Science Inc. Woodland Hills, CA, USA). The diameter of the rotating rod was 3.18 cm. A first "trial run" was performed, prior to the experimental test. Once all animals were on the rod, the motor was turned on at a starting speed of four RPM, and the rod rotation was continuously accelerated until it reached a maximum speed of 40 RPM after 300 s, which was the cut-off point to end the test. For each mouse, the trial was ended when the mouse fell from the rotating rod, or at the end of the cut-off point. The time it took the animal to fall from the rotating rod was recorded by a blinded experimenter.

Y-Maze
The Y-maze assesses spontaneous alternation as a readout for exploratory behavior and working memory in rodents. Our Y-maze consisted of 3 identical arms at 120° angles from each other. Mice were introduced to the test room 10 min or more prior to testing. The test room, lighting, and position of the Y-maze were the same each time. The paper floor of the maze was replaced fresh and wiped clean for each trial. The mouse was carefully placed into the center of the Y-maze and allowed to explore freely for 5 minutes. Each complete entry into one arm of the Y-maze labeled A, B, or C was scored manually if all 4 limbs crossed the threshold. Upon completion, each mouse was returned carefully to their respective cage. Total entries and triads were recorded for each mouse to determine changes in treated vs control. Mice were tested before and after treatments and fold changes calculated accordingly.

Morris Water Maze
The mice were assessed post-injection with the Morris water maze for spatial learning and memory. The water maze was a steel circular pool 152 cm in diameter. It was filled with opaque white water at 21-22°C. A black, square platform (12.7 cm x 12.7 cm) was used for the training. Different shaped cutouts were placed just above the tank's rim in each quadrant to be used as visual cues. Prior to the training sessions, the mice were handled for the first 3 days within the same room as the water maze to acclimate them to the handlers and the room. On day 4-6, the mice were given 6 sessions to train with a platform that was raised 0.2 cm above the water level. During the visible platform training, both the platform and the drop location changed for each session. On day 7-11, there were 10 sessions with a hidden platform by submerging it 1.5 cm below the water level.
Throughout all the hidden platform sessions, the platform remained stationary while the drop location was changed. On day 12, a probe test without the platform was performed for 2 minutes. The tests were recorded and analyzed for the time spent within each quadrant.

In Vivo Scoring
Animals were scored for the onset of symptoms including dystonia, coat condition, ataxia, body condition, gait disorders. The investigator recording onset and severity of symptoms was blinded as to treatment groups. All symptoms were scored on a scale of 0 to 10, with: a) 10 indicating no symptom; b) 5 indicating the symptom was detected but not consistently expressed; c) 0 indicating when the animal begins to express the symptom 100% of the time. For fur conditions, a scoring was assigned for each of the following conditions: 1) alopecia; 2) dermatitis; 3) loss of pigmentation; 4) lucidity; 5) organization.

Ex vivo force production
An 800A in vitro apparatus (Aurora Scientific) was used to perform the ex vivo testing of the TA muscles. The testing chamber, along with the Krebs solution (Sigma), was kept at 25 °C using an external water heater. To help ensure tissue viability, a 95% O2, 5% CO2 mixture was used to oxygenate the testing chamber along with a petri dish containing Krebs solution. Mice were killed and the lower limbs excised from the ankle to B0.5 inches above the patellar tendon to ensure the entire TA muscle was removed. The legs were then pinned down in the petri dish coated with an elastomer (Sylgard-184, Dow Corning).
The TA muscles were carefully dissected from the legs under a microscope and held under tension until attached to the force transducer lever. The muscle was attached to the force transducer lever by directly tying a knot around the top of the patellar tendon. The TA muscle was then attached to the force transducer lever and the bottom of the testing apparatus to secure the TA muscle in place. The testing chamber was raised around it to immerse the muscle in Krebs solution. A similar testing procedure to the in vivo protocol was used ex vivo except no voltage recruitment testing was performed. Imaging was performed in bind: the investigators performing the imaging did not know the identity of the experimental conditions for the transplanted cells.
Immediately following the injection, mice were imaged for 30 s at maximum sensitivity (fstop 1) at the highest resolution (small binning). Every minute a 30 s exposure was taken, until the peak intensity of the bioluminescent signal began to diminish. Each image was saved for subsequent analysis. Imaging was performed in bind: the investigators performing the imaging did not know the identity of the experimental conditions for the transplanted cells.

Bioluminescence image analysis
Analysis of each image was performed using Living Image Software, version 4.0 (Caliper Life Sciences). A manually-generated circle was placed on top of the region of interest and resized to completely surround the limb or the specified region on the recipient mouse.
Similarly, a background region of interest was placed on a region of a mouse outside the transplanted leg.

Histology
TA muscles were fixed for 5 h using 0.5% electron-microscopy-grade paraformaldehyde and subsequently transferred to 20% sucrose overnight. Muscles were then frozen in optimal cutting temperature, cryosectioned at a thickness of 10 mm and stained. For colorimetric staining with Hematoxylin and Eosin (Sigma) or Gomorri Trichrome (Richard-Allan Scientific) samples were processed according to the manufacturer's recommended protocols. For SA-β-Gal, stock solution was prepared using Corp X-GAL (Fisher Scientific) and staining was performed in combination with Feulgen Stain Kit (American Mastertech) for nuclear staining.

Immunostaining
A 1 h blocking step with 20% donkey serum/0.3% Triton in PBS was used to prevent unwanted primary antibody binding for all samples. Primary antibodies were applied and allowed to incubate overnight at 4°C in 20% donkey serum/0.3% Triton in PBS. After four washes with 0.3% PBST, fluorescently conjugated secondary antibodies were added and incubated at room temperature for 1 h in 0.3% PBST. After three additional rinses each slide was mounted using Fluoview mounting media. Photographs were acquired using standard fluorescent microscope and a CCD camera. Volocity imaging software. Was used to adjust excitation and emission filters and came with AlexaFluor filter settings, which were used whenever possible. All exposure times were optimized during the first round of imaging and then kept constant through all subsequent imaging. The immunoreactivity was quantified as the percent area covered by ImageJ (ref). For each staining, a total of three hippocampal brain sections per mouse were analyzed.

Muscle Stem Cell Isolation and purification
Muscles were harvested from hind limbs and mechanically dissociated to yield a fragmented muscle suspension using a gentleMACS dissociator (Miltenyl Biotec). This was followed by a 90 min digestion in a Collagenase II-Ham's F10 solution (500 units per ml; Invitrogen). After washing, a second digestion was performed for 30 min with Collagenase II (100 units per ml) and Dispase (2 units per ml; Invitrogen). For Pax7CreER/ROSA26eYFP mice, MuSCs were directly sorted by endogenous eYFP protein expression. Cell sorting was performed on calibrated BD-FACS Aria II or BD FACSAria III flow cytometers equipped with 488-nm, 633-nm and 405-nm lasers to obtain the MuSC population. A small fraction of sorted cells was plated and stained for Pax7 and MyoD to assess the purity of the sorted population

Myogenic colony-forming cell assay for MuSCs
Single treated and control MuSCs were deposited into wells of collagen-and laminincoated plates at 1 cell per well by BD FACSAria III flow cytometer. Collagen/laminin coating was accomplished by overnight incubation of the plates rocking at 4°C with a 1:1 mixture of laminin (10 μg/ml ThermoFisher 23017-015) and collagen (10 μg/ml Sigma C8919) in PBS. Coated wells were washed three times with PBS before use. The cells were cultured in grow media, F10 medium supplemented with 20% horse serum and 5 ng/ml basic fibroblast growth factor (bFGF; PeproTech 100-18B). After 6 days of culture, plates were fixed with 4% paraformaldehyde (Electron Microscopy Services 15710), stained with DAPI (Invitrogen D1306), and scored by microscopy to determine the number of myogenic colony-forming cells (MCFCs), defined by wells that contained at least 8 cells.

EDU staining
Staining was done according to the manufacturer protocol using the Click-iT EdU kit. Cells were label with Edu after switching to growth media. Cells were allowed to grow one day before fixation with 4% paraformaldehyde and permeabilization with 0.5% Triton X-100 in PBST. Cells were the incubated in Click-It reaction cocktail for 30 min before washing in PBS and imaging.

Complete Blood Counts
Mice were anesthetized using isoflurane and blood was collected via cardiac puncture using a 1mL syringe with a 25 gauge needle. Blood was transferred immediately to a K2-EDTA microtainer tube, inverted 8 to 10 times and placed it on a rocker at room temperature. Blood was tested on a VETSCAN HM5 Hematology Analyzer running a software VetScan HM5 v2.3.

Quantitative PCR
For RNA extraction, samples were collected in RNAlater solution (Sigma-Aldrich, USA).

Image analysis
All other analyses were performed using Image J or Volocity software and manually counting fibres using the free draw tool and also counting the number of nuclei and CSA of myofibers.

Data availability
The data that support the findings of this study are available from the corresponding author upon reasonable request.