Therapeutic targeting of senescent cells in the CNS

Abstract

Senescent cells accumulate throughout the body with advanced age, diseases and chronic conditions. They negatively impact health and function of multiple systems, including the central nervous system (CNS). Therapies that target senescent cells, broadly referred to as senotherapeutics, recently emerged as potentially important treatment strategies for the CNS. Promising therapeutic approaches involve clearing senescent cells by disarming their pro-survival pathways with ‘senolytics’; or dampening their toxic senescence-associated secretory phenotype (SASP) using ‘senomorphics’. Following the pioneering discovery of first-generation senolytics dasatinib and quercetin, dozens of additional therapies have been identified, and several promising targets are under investigation. Although potentially transformative, senotherapies are still in early stages and require thorough testing to ensure reliable target engagement, specificity, safety and efficacy. The limited brain penetrance and potential toxic side effects of CNS-acting senotherapeutics pose challenges for drug development and translation to the clinic. This Review assesses the potential impact of senotherapeutics for neurological conditions by summarizing preclinical evidence, innovative methods for target and biomarker identification, academic and industry drug development pipelines and progress in clinical trials.

Introduction

Cellular senescence was first described in 1961 as a phenomenon whereby primary cells in culture undergo a finite number of divisions¹. This limited replicative potential was believed to be an artefact of cells in culture driven by telomere erosion^2,3. The eventual discovery of molecular biomarkers enabled the identification of senescent cells in vivo^4,5. However, the inclusion of cell cycle arrest in the definition of senescence cast doubt on the relevance of senescence in tissues with mostly non-dividing cells, such as the brain. The first study to report senescent neurons in mouse brain tissue was transformative to the field⁶ but was met with scepticism. Later discoveries of senescent-like neurons^7–10 and astrocytes^11,12 in post-mortem tissue from patients with Alzheimer disease (AD) and Parkinson disease supported the initial findings and highlighted the potential translational importance of senescent cells in the brain. These studies, along with the identification of postmitotic senescent-like cells in other tissues¹³, inspired the field to reconsider the senescence-defining criteria. Cellular senescence now refers to an end-stage change in cell fate, orchestrated through simultaneous and sustained activation of opposing pro-apoptotic and pro-survival molecular pathways¹⁴. For the context of this Review, we focus only on pathological senescence that arises through chronic stress activation, distinguishing it from physiological senescence, which occurs during development, tissue regeneration or wound healing^15,16.

Similar to other changes in cell fate, senescent cells acquire distinct morphologies and remain metabolically active¹⁷. They often develop enlarged somas both in vitro¹⁸ and in vivo^9,19, exhibit increased lysosomal senescence-associated β-galactosidase (SA-β-gal) activity^20,21, elevated glycolysis²² and lipid metabolism²³, and acquire a secretory phenotype^16,24. The molecules they release, known as the senescence-messaging secretome (SMS)²⁴, or more commonly as the senescence-associated secretory phenotype (SASP)¹⁶, communicate with their neighbouring environment and the immune system either to prepare the tissue for disease or to facilitate their own clearance, respectively. The presence of senescent cells perpetuates tissue inflammation and destruction through multiple mechanisms²⁵. If senescent cells are not actively removed by the immune system, their sustained SASP signalling induces cytotoxicity or transforms healthy cells into senescent cells. Phagocytic cells can also become senescent after clearing senescent cells, which perpetuates inflammation and pathogenic processes (Fig. 1). Throughout the lifespan, tissues and organs accumulate senescent cells, and their contribution to disease and dysfunction earned them recognition as a “biological hallmark of aging”²⁶.

Fig. 1 |. Senescence initiation and spread across brain cells.

a, Age and disease-associated stressors that drive inflammation, reactive oxygen species, toxic protein accumulation, metabolic dysregulation or DNA damage may cause senescence in many cell types, including neurons. b, Senescent neurons exhibit altered excitability and/or activity, display ‘eat me’ signals and secrete deleterious molecules (the senescence-associated secretory phenotype (SASP)) that negatively impact neuronal, vascular and glial cells. They may also contain aggregate-prone, neurotoxic proteins that they transmit to other cells as neuronal SASP^9,19. c, Activated microglia expressing phagocytic receptors recognize neuronal phagoptosis ‘eat me’ signals. d, Activated microglia engulf senescent neurons¹⁶⁶ and their content, including difficult-to-digest protein aggregates, which may cause microglial senescence. e, Senescent microglia exhibit reduced phagocytic and surveillance function⁴⁴. They release partially digested, neurotoxic fragments of protein aggregates and SASP factors that cause astrocytes⁴², microglia^44,57, vasculature⁴³ and oligodendrocyte precursor cells⁴⁵ to become senescent. Aβ, amyloid-β; Htt, huntingtin; PrP, prion protein; PS, phosphatidylserine; SOD1, superoxide dismutase 1; TDP43, TAR DNA-binding protein 43.

During the transition to senescence, cells undergo significant changes in chromatin structure, alter their metabolism and morphology, and become resistant to apoptosis¹⁴. The features of senescent cells depend on the original cell type and the specific stressor involved²⁷. Various markers have been used to identify senescent cells, but a confident identification requires evidence of multiple overlapping phenotypes within the same cell, such as changes in morphology, macromolecular damage, cell cycle arrest and an inflammatory secretome^14,28. In addition to applying traditional senescence detection methods to tissues of interest, the advancement in single-cell and single-nucleus RNA sequencing, high-resolution multiplex spatial profiling techniques and machine learning (ML) and artificial intelligence (AI) enable the identification of rare senescent cells in complex tissues^19,29–32. These strategies have propelled the discovery of novel biomarkers of senescence. Experimental evidence with new and established biomarkers indicates that several brain cell types, including neurons, glia, stem cells and cerebral vasculature cells, may undergo senescence owing to a range of stressors, such as physical, psychological, viral, oxidative and proteotoxic insults, DNA damage and protein accumulation (Table 1).

Table 1 |.

Evidence of cellular senescence in ageing and CNS conditions

Condition	Astrocytes	Microglia	Neurons	Other CNS cells
Alzheimer disease	p16 (refs. 11,251), p53 (ref. 251)	p16 (ref. 202) morphology, DNA damage, HMGB1 (ref. 252); protein panel (ref. 251)	Transcriptome (refs. 9,19,208,253) p19+ NFTs, karyomegaly, lipofuscin (ref. 19)	OPCs (ref. 45) Endothelial cells (ref. 200)
Ageing	p16 (refs. 11,254); p21, SA-β-gal254; loss of HMGB1, SASP, p16 (ref. 42)	p16 (ref. 255); protein panel (ref. 251)	p21 (ref. 6); Gdf11 knockout (ref. 104)	OPCs (ref. 255) VSMCs (ref. 31) Neuroblasts (ref. 256)
Amyotrophic lateral sclerosis	SA-β-gal (ref. 257), DNA damage (ref. 257) p16 (refs. 251,258–260), p53 (ref. 251) p21 (ref. 258); γH2AX (ref. 257)	p16 (refs. 259,260), loss of lamin B1 (ref. 259)	p16 (ref. 259); p21 (ref. 258)	–
Chronic pain	Telomere erosion, p53 (ref. 53)	Telomere erosion, p53 (ref. 53) SA-β-gala (ref. 261)	Telomere erosion, p53 (ref. 53)	–
Parkinson disease	Loss of lamin B1 (ref. 12)	–	p21 and loss of SATB1 (ref. 10)	–
Stroke	SA-β-gal (ref. 262), p16 (refs. 262,263); SASP (ref. 263)	p16, p21, SASP (ref. 264)	p16, p21, SASP (ref. 264)	Endothelial cells (ref. 263)
Tauopathy	FTLD: HMGB1, SASP, p16 (ref. 42)	SA-β-gal, SASP, MMP3 (ref. 44); p16 (ref. 57)	PSP: Cdkn2a, DNA damage, karyomegaly (ref. 9)	–
Traumatic brain injury or blast exposure	Cyclin D1, PCNA, p16, p21, SA-β-gal (ref. 265); γH2AX, loss of lamin B1, SASP (ref. 266)	Cyclin D1 (ref. 265); PCNA (ref. 265); p16 (refs. 265,267); p21 (refs. 265,267); SA-β-gal (ref. 265); Bcl-2 (ref. 267), lipofuscin (ref. 267) and γH2AX (ref. 267)	DNA damage (ref. 51); p16 (refs. 51,265); p21 (refs. 51; 265) Cyclin D1, PCNA, SA-β-gal (ref. 265)	Ependymal cells (ref. 266) Oligodendrocytes (ref. 266)
Other conditions	Obesity (ref. 49)	–	Alcohol use disorder (ref. 48); Glaucoma (ref. 268) Insulin resistance (ref. 50) Spinal cord injury (ref. 52) SARS-CoV-2 (ref. 269)	Glioblastoma (refs. 203,270,271) MS: NPCs (ref. 272)

CNS, central nervous system; FTLD, frontotemporal lobar degeneration; MS, multiple sclerosis; NFT, neurofibrillary tangle; NPC, neural precursor cells; OPC, oligodendrocyte precursor cell; p16, p16^INK4a; p19, p19^INK4d; p21, p21^CIP1; PSP, progressive supranuclear palsy; S-β-gal, senescence-associated β-galactosidase; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2; SASP, senescence-associated secretory phenotype; VSMC, vascular smooth muscle cell. Evidence of senescence using bulk tissues without cell specificity not included.

^aCell type inferred from in vitro model, but not confirmed in vivo.

In this Review, we describe mechanisms involved in central nervous system (CNS) cell senescence, and consider how these processes might be therapeutically targeted. We balance optimistic promise for their potential, including results from first clinical trials³³, with an objective perspective on the challenges faced and strategies to overcome them.

Mechanisms leading to CNS cell senescence

Cell cycle dysfunction^34,35, ‘loss of cell identity’^36,37 or lysosomal stress³⁸ may all represent key steps in the conversion from a healthy brain cell type into a final senescent state. Neurons, the cellular building blocks of the nervous system, communicate through electrical signals. They are necessary for the processing, storage and retrieval of information. Neurons are vulnerable to somatic mutations^39,40 and insoluble protein accumulation⁴¹, which are known drivers of senescence through DNA damage pathway activation¹⁰ and protein aggregation^9,19.

Non-neuronal brain cell types, such as glia, immune cells, endothelial cells and vascular smooth muscle cells, have crucial roles in maintaining the health and function of the nervous system by modulating neuronal activity, preventing and removing toxins, and nourishing the tissue. These cells may be especially vulnerable to cell non-autonomous stressors such as external protein aggregates and cellular by-products. For example, in AD and other tauopathies, intraneuronal tau aggregates drive neuronal senescence^9,19, and tau oligomers released from neurons trigger astrocyte⁴² and vascular senescence⁴³. Microglia can enter senescence after phagocytosing protein aggregates, including clearing senescent neurons that contain neurofibrillary tangles (NFTs) composed of aberrantly phosphorylated tau aggregates⁴⁴. Similarly, amyloid-β, another hallmark of AD, can induce senescence in oligodendrocyte precursor cells⁴⁵ and endothelial cells⁴⁶. TNF, a cytokine secreted by many cell types, including brain microglia and senescent cells, can induce neuronal senescence, increasing the release of α-synuclein⁴⁷, a synaptic protein that aberrantly accumulates in Parkinson disease and other neurodegenerative diseases. A model depicting senescent cell initiation and propagation across brain cell types is presented in Fig. 1, and the features of senescent cells, including postmitotic senescent cells, are depicted in Fig. 2.

Fig. 2 |. Cellular phenotypes of postmitotic senescent neurons.

Cellular phenotypes of postmitotic senescent neurons include morphological changes, lysosomal and mitochondrial dysfunction, DNA damage, increased nuclear size, decreased expression of lamin B1, increased expression of p21, p16 and p19, changes in membrane potential and a senescence-associated secretory phenotype (SASP). The black arrows indicate axonal and dendritic retraction, which is a feature of postmitotic senescent neurons that does not occur in non-neuronal cell types. Processes highlighted in blue are being explored as senotherapeutic targets across senescent cell types. Increased β-galactosidase, lipofuscin accumulation, proliferation arrest and telomere attrition are senescence phenotypes that may occur in replicative or stress-induced senescence but are not specific markers for postmitotic senescent neurons and are not shown. p16: p16^INK4a; p19: p19^INK4d; p21: p21^CIP1.

Senescent cells that accumulate within the brain or in peripheral tissues negatively impact brain health and function (Table 1). Senescent cell accumulation in the brain may be driven by comorbidities that increase the risk of developing cognitive impairment or brain dysfunction (for example, insulin resistance, obesity, alcohol use disorder^48–50), or by physical stress to the nervous system such as traumatic brain injury⁵¹ or spinal cord injury⁵². Emerging evidence suggests that senescence is a contributing and causal factor for chronic pain and psychological stress-induced mental disorders, primarily through the release of inflammatory SASP factors^49,50,53,54. Individuals with a high blood SASP index tend to experience more severe depressive episodes than individuals with low SASP indices⁵⁵. Furthermore, elevated SASP expression levels have been linked to poorer treatment response and reduced remission rates in late-life depression⁵⁶. These findings in humans are supported by mechanistic studies in mice, demonstrating a causal link between senescence and depressive behaviours⁴⁹.

In 2018, studies began to report improvements in brain structure and function by elimination of senescent cells in transgenic mouse models of neurodegeneration ^9,12,57,58. Mechanistic data indicated that clearing senescent cells improved brain structure, function and memory and reduced neuroinflammation and AD-associated neuropathology (Table 2). The clear link between senescent cell accumulation and organ dysfunction and healthy lifespan^59–62, along with the potential for therapeutic interventions, led to a surge of funding to support research on cellular senescence in ageing and disease, including in the CNS^63,64.

Table 2 |.

Selected senotherapeutic molecular targets under preclinical investigation for CNS conditions

Drug target	Compounds (examples)	BBB penetrance	Condition/indication (CNS-specific only)	Mechanism
SRC kinase plus PI3K	Dasatinib–quercetin (ref. 65)	Dasatinib: yes (ref. 273) Quercetin: yes/weak (refs. 274,275)	Parkinson disease mice (ref. 12) Overtraining-induced deficits in Learning and memory (ref. 276) Age-related cognitive decline (ref. 277) Tauopathy mice (ref. 9) AD amyLoid mice (ref. 45)	Senolysis: activation of apoptosis in senescent cells
mTOR	Rapamycin (ref. 148)	Yes/weak (ref. 278)	Brain health/ageing/memory (refs. 279,280)	Blocks NRF2-mediated senescence induction and reduces NF-κB-mediated SASP gene expression
HSP90	Geldanamycin (ref. 101)	No (ref. 281)	ND	Autophagy-mediated senolysis
	Tanespimycin (ref. 101)	Yes (ref. 282)	ND
	Alvespimycin (ref. 101)	Yes (ref. 282)	ND
	Ansamycin (ref. 101)	ND	ND
	Resorcinol (ref. 101)	ND	ND
	Purine and pyrimidine-like N-terminal HSP90 inhibitors (ref. 101)	ND	ND
Seno-antigens: Surface markers of senescent cells	uPAR targeted by CAR-T cells (ref. 169)	Yes (ref. 283)	ND	Phagocytosis-mediated clearance of senescent cells
	Antibody-based approach for DPP4 (ref. 170)	No (ref. 284)	ND
	GPNMB as immunization antigen (ref. 171)	No (ref. 284)	ND
Na+K+-ATPase	Proscillaridin A (ref. 103)	ND	ND	Senolysis by modulating plasma membrane potential
	Ouabain (refs. 103,105)	Yes (ref. 285)	TBI (ref. 289)
	Bufalin (refs. 103,105)	Yes (ref. 286)	ND
	Cinobufagin (ref. 103)	ND	ND
	Peruvoside (ref. 103)	ND	ND
	Digitoxin (refs. 103,105)	Weak (ref. 287)	ND
	Convallatoxin (ref. 103)	Yes (ref. 288)	ND
	Digoxin (refs. 103,105)	Weak (ref. 287)	Dementia (ref. 290)
BCL-2 protein Family	ABT-737 (ref. 107)	No/weak (ref. 292)	COVID-19 neuropathology (ref. 294)	Senolysis: deactivation of anti-apoptotic pathway/activation of apoptosis in senescent cells
	ABT-263 (navitoclax) (refs. 61,90,109)	No/weak (ref. 292)	BBB disruption (ref. 295) AD (ref. 57) COVID-19 neuropathoLogy (ref. 294) Learning and memory in age (ref. 296)
	A-1331852 (ref. 90)	Yes (ref. 293)	ND
	A-1155463 (ref. 90)	ND	ND
	UBX1325 (ref. 291)	ND	ND
PUMA/NOXA	PCC1 upregulates PUMA/NOXA (ref. 139)	ND	ND	Mitochondria-mediated senolysis
USP7	P5091 (ref. 121)	Modified (ref. 297)	Tau phosphoryLation (AD modeL) (ref. 299) ND	p53-mediated senolysis (elevated p53 expression)
	P22077 (ref. 121)	Yes (ref. 298)
MDM2	Nutlin3a (ref. 300) and UBX1325 (ref. 223) attenuate SASP UBX0101 (ref. 301) RG-7112 (ref. 302)	No/weak (ref. 303), unclear	ND	p53-mediated senoLysis (eLevated p53 expression)
			ND
		ND	ND
		Yes (ref. 304)	ND
p53–FOXO4 Interaction	FOXO4-DRI (cell-penetrating peptide that disrupts interaction of p53 with FOXO4) (ref. 120)	ND	ND	p53-mediated senoLysis (p53 cytosolic translocation)
HDAC	Panobinostat (ref. 129)	Yes (ref. 305)	ND	BCL-XL and H3-mediated senolysis
NF-κB	BAY11-7082 (ref. 306)	ND	Aggregation of amyloid-β protein and memory deficits (ref. 307)	Reduced NF-κB activity to decrease SASP gene expression
BRD4	ARV825 (ref. 132)	Modified (ref. 308)	ND	PROTAC, senolysis mediated by elevated DNA damage
IKK	EF24 (curcumin analogue) (ref. 309)	ND	ND	Reduced NF-κB activity to decrease SASP gene expression
ATM	KU-55933 suppressed senescence and SASP (ref. 143)	No (ref. 310)	Neuroprotection against hydrogen peroxide-induced cell damage (ref. 311)	Reduced NF-κB activity to decrease SASP gene expression
OXR1	Piperlongumine (ref. 135)	Yes (ref. 312)	Age-related cognitive impairment (ref. 313)	Induces degradation of OXR1 to induce senolysis
Lysosome	Cytotoxic drugs encapsulated with galacto-oligosaccharides (ref. 114) e.g., Nav-Gal (ref. 116), 5FURGal (ref. 314), SSK1 (ref. 315)	ND	ND	Cytotoxic drugs are released in the senescent lysosomes, causing senolysis
SASP	SR12343 (ref. 154)	Yes (ref. 155)	Improved dystonia and motor function (ref. 154)	Blocks NF-κB-IKK; reduces NF-κB-mediated SASP gene expression p38 inhibition; reduces NF-κB-mediated SASP gene expression Reduced C/EBPβ activity leading to lower expression of SASP genes
	SB203580 (ref. 158)	ND	ND
	UR-13756 (ref. 159)	ND	ND
	BIRB 796 (ref. 159)	ND	ND
	MW01-18-150SRM (ref. 316)	Yes (ref. 317)	Neuroinflammation, psychiatric and cognitive behaviour (AD) (refs. 316,319)
	Ruxolitinib (suppressor of C/EBPβ activity) (ref. 161)	Yes (ref. 318)	Learning and memory (ref. 320) TBI (ref. 321)
BCL-XL	PZ15227 (ref. 113)	ND	ND	ABT-263-based BCL-XL PROTAC; senolysis
cGAS	Aspirin	Yes (ref. 322)	Reduced SASP and senescence in brain organoids (ref. 182)	Reduced SASP gene expression by inhibition of COX2
STING	H-151	Yes (ref. 183)	Ameliorated SASP in aged mice (ref. 183) and reduced senescence in brain organoids (ref. 182)	Reduced SASP gene expression by reducing NF-κB activity

AD, Alzheimer disease; ATM, ataxia–telangiectasia mutated; BBB, blood–brain barrier; CAR, chimeric antigen receptor; CNS, central nervous system; COVID-19, coronavirus disease 2019; COX2, cyclooxygenase 2; DPP4, dipeptidyl peptidase 4; FOXO4, forkhead box protein O4; GPNMB, glycoprotein nonmetastatic melanoma protein B; HDAC, histone deacetylase; HSP90, heat shock protein 90; IKK, IκB kinase; mTOR, mechanistic target of rapamycin; ND, not determined; OXR1, oxidation resistance gene 1; PCC1, procyanidin C1; PI3K, phosphatidylinositol 3-kinase; PROTAC, proteolysis-targeting chimera; SASP; senescence-associated secretory phenotype; TBI, traumatic brain injury; uPAR, urokinase-type plasminogen activator receptor; USP7, ubiquitin-specific peptidase 7.

The pioneering discovery of compounds that could eliminate senescent cells, including in the brain, stimulated great interest in drug development⁶⁵. Therapeutic approaches under investigation include methods to clear senescent cells from tissue, to gently revert them back to their original cell type or to dampen their SASP. The type of senescent cell and its associated mechanism of toxicity influence the choice of therapeutic strategy. For instance, drugs designed to clear senescent cells may be most suitable for replaceable cell types, such as microglia; whereas non-clearing strategies may be more appropriate to target cells that are difficult to replace, such as neurons. Research into lifestyle interventions and other non-pharmacological strategies to reduce senescent cell burden, including approaches such as mindfulness to treat insomnia in AD caregivers (NCT03538574), is also active and promising, and may be effective for some individuals^66–68. However, this Review focuses on the various pharmacological approaches under investigation (Box 1).

Box 1 |. Senescent cell-targeting strategies.

Senotherapeutic drug strategies broadly fall into two categories, senolytics and senomorphics. Senolytics clear senescent cells from the tissue, whereas senomorphics allow senescent cells to remain but mitigate their pathogenic properties. Non-pharmacological strategies are also gaining attention, but are beyond the scope of this Review.

Senolytics

• Senescence-associated pathway inhibition

  • SRC kinase and phosphatidylinositol 3-kinase inhibitors
  • Heat shock protein 90 inhibitors
  • Na⁺K⁺-ATPase inhibitors
  • BCL-2 protein family inhibitors
  • Ubiquitin-specific peptidase 7 (USP7), MDM2 and p53–forkhead box protein O4 (FOXO4) interaction inhibitors
  • Histone deacetylase inhibitors
  • BRD4 inhibitors
• Harnessing the immune system

  • Glycoprotein nonmetastatic melanoma protein B vaccine
  • Chimeric antigen receptor-T therapy (urokinase-type plasminogen activator receptor (uPAR))
• Prodrugs

  • Galactose-modified duocarmycin prodrugs
• Proteolysis-targeting chimeras

  • PZ15227
  • ARV825

Senomorphics

• Senescence-associated secretory phenotype modulation

  • Rapamycin
  • Metformin
  • NF-κB inhibitors
  • p38 inhibitors
  • JAK–STAT inhibitors
  • Ataxia–telangiectasia mutated (ATM) inhibitors
• Cell reprogramming

  • CDGSH iron–sulfur domain 2
  • G-quadruplex structures
  • DNA methylation
  • Mitochondrial decay

Non-pharmacological senotherapeutics

• Caloric restriction

• Intermittent fasting

• Exercise

CNS senotherapeutics in development

To develop effective senotherapeutics for the CNS, strategies need to incorporate various drug modalities against diverse targets and senescent cell types. This section describes therapeutic strategies that range from inhibition of specific senescence-associated signalling pathways, such as those that control SASP, to activation of apoptosis-inducing pathways. Several molecules have been identified as potential senotherapeutic targets (Table 2 and Fig. 3). Small-molecule approaches are the most straightforward option for targeting senescent cells in the CNS owing to their potential for oral bioavailability and relatively favourable profile regarding CNS penetration across the blood–brain barrier (BBB). Candidate molecular processes can be targeted with senolytics, which destroy senescent cells, or with senomorphics, which modulate senescent cells (Box 1). We also explore innovative approaches such as targeting proteins specific to senescent cells for degradation. Although not all the targets and approaches discussed in the following paragraphs have been directly tested in the CNS, they describe basic principles for tackling senescent cells and lay the groundwork for the future development of CNS-targeted applications.

Fig. 3 |. Potential molecular drug targets for CNS senotherapeutics.

Senescent cells feature upregulated pro-survival pathways and a senescence-associated secretory phenotype (SASP). Several molecular players in these pathways are under investigation as potential targets for senotherapeutic strategies. Drug-targeting approaches to activate (red arrows), inhibit (red inhibitory lines) or promote (dashed arrows) protein transport within the cell can result in cell clearance (senolytics) or modulate the SASP (senomorphics). A combination of dasatinib and quercetin, which target Src tyrosine kinase signalling, were the first generation of potential senolytics to be discovered. Other promising approaches include targeted inhibition of the pro-survival BCL-2-related pathways, or NF-κB signalling to ameliorate the expression of SASP genes. Selected compounds are shown as examples (red boxes); see Table 2 for further details. Note that galacto-oligosaccharide (gal)-encapsulated drugs, which target raised β-galactosidase levels in the lysosomes of senescent cells, might not be suitable for central nervous system (CNS) applications owing to high β-galactosidase expression in non-senescent neurons. COX2, cyclooxygenase 2; DPP4, dipeptidyl peptidase 4; GPNMB, glycoprotein nonmetastatic melanoma protein B; HDAC, histone deacetylase; HSP90, heat shock protein 90; OXR1, oxidation resistance gene 1; RTK, receptor tyrosine kinase; Ub, ubiquitin; uPAR, urokinase-type plasminogen activator receptor; USP7, ubiquitin-specific peptidase 7.

Senolytics

Senolytics are a class of small-molecule therapies designed to selectively eliminate senescent cells. The term ‘senolytics’ originates from combining ‘senescence’ and ‘lytic,’ indicating the literal destruction of senescent cells. These cells are inherently damaged with upregulated apoptotic pathways. Their long-term survival relies on concurrent upregulation of pro-survival pathways inhibiting full engagement of apoptosis. Many senolytics function by deactivating these survival pathways, referred to as senescent cell anti-apoptotic pathways (SCAPs), allowing senescent cells to undergo apoptosis (Table 2 and Fig. 3). The therapeutic druggability of senescence survival pathways is explored further below.

PI3K and Src kinase inhibition.

Phosphatidylinositol 3-kinases (PI3Ks) are lipid kinases that have been shown to take crucial roles in the regulation of cell cycle, apoptosis, DNA repair, angiogenesis, cellular metabolism and cellular senescence⁶⁹. Dysregulation of the PI3K pathway is observed in one-third of human tumours, which has led to the development of PI3K inhibitors as cancer therapeutics⁷⁰. Similarly, the Src family kinases are involved in regulation of important cellular functions such as cell proliferation, differentiation, apoptosis, migration and metabolism⁷¹. Early activation of Src has been implicated in shifting the cell fate from apoptosis to senescence⁷².

A significant advancement in the field of senolytic agents was the discovery of a combinatorial therapeutic approach involving dasatinib and quercetin. Dasatinib, a Src tyrosine kinase inhibitor with FDA approval for leukaemia treatment, was combined with quercetin, a naturally occurring flavonoid that inhibits PI3K⁶⁵. Although dasatinib had already demonstrated its ability to induce apoptosis in cancer cells, it initially showed no significant effects on senescent cell clearance when used alone. Similarly, quercetin, a relatively nonspecific PI3K inhibitor known for its anti-inflammatory properties and broad kinase inhibitory profile, demonstrated only modest senolytic potential as a monotherapy. However, the combination of dasatinib with quercetin synergistically influenced SCAPs and demonstrated significant senolytic potential⁶⁵.

This combination of dasatinib and quercetin has since proved effective in eradicating senescent cells and improving health outcomes in mouse models of tau-associated neurodegeneration⁹, AD-associated amyloid-β accumulation⁴⁵, frailty⁶⁵, osteoporosis⁶⁵, vasomotor dysregulation^65,73, hepatic steatosis⁷⁴, insulin resistance⁷⁵, pulmonary fibrosis⁷⁶, adipose tissue inflammation⁷⁷, intervertebral disc degeneration⁷⁸ and skeletal muscle debility⁷⁹. The success of dasatinib–quercetin has stimulated considerable interest in combatting age-related CNS conditions in human trials (Table 3). Although the side effects of long-term use of dasatinib–quercetin remain unknown, to date all the clinical trials have reported favourable safety outcomes. Chronic complications of dasatinib treatment in cancer therapy include recurrent pleural effusions in up to 37% of patients, as well as possible side effects affecting platelet function, leading to a bleeding tendency⁸⁰. To minimize potential severe side effects such as these, dasatinib–quercetin senolysis treatment is administered intermittently. Despite dasatinib’s short elimination half-life of less than 11 h, and quercetin’s poor bioavailbility⁸¹, early data from clinical trials indicate that dasatinib–quercetin is potent enough to significantly reduce the senescent cell burden in humans using this ‘hit-and-run’ regimen⁸², which simultaneously improves the safety profile.

Table 3 |.

Selected senotherapeutics under clinical investigation for central nervous system conditions

Compounds	Targets	Indication	Sponsor	Study phase/type	NCT identifier
Dasatinib–quercetin combination therapy	Primary target Dasatinib: SRC kinase (ref. 65) Quercetin: PI3K (ref. 65)	MCI with slow gait speed	Hebrew Senior Life	Phase I/II open label	NCT05422885
	Other drug targets Dasatinib: ABL kinases (ref. 323) (ABL1, BCR–ABL); receptor tyrosine kinases (ref. 324) (EGFR, EPHA2, DDR1); non-receptor tyrosine kinases (ref. 324) (FRK, BRK, ACK), p90RSK (ref. 324) Quercetin: suppression of COX2, NF-κB and AP-1 pathways (ref. 325), nonspecific, multiple targets (ref. 326)	MCI/early AD	University of Texas Health Science Center at San Antonio	Phase I/II POC open label	NCT04063124
		MCI/early AD	Wake Forest University Health Sciences	Phase II RCT	NCT04685590
		MCI/early AD	Mayo Clinic	Phase I/II open label	NCT04785300
		Older adults with depression or schizophrenia	Washington University School of Medicine	Phase II RCT	NCT05838560
UBX1325	BCL-XL	DME or neovascular AMD	Unity Biotechnology	Phase I open label	NCT04537884
		Wet AMD	Unity Biotechnology	Phase II RCT	NCT05275205
		DME	Unity Biotechnology	Phase IIb	NCT06011798

Non-pharmacological trials or those using compounds with senotherapeutic potential (e.g., polyphenols, rapamycin, etc.) but not in a senescence indication are not included. AD, Alzheimer disease; AMD, age-related macular degeneration; DME, diabetic macular oedema; MCI, mild cognitive impairment; PI3K, phosphatidylinositol 3-kinase; POC, proof-of-concept; RCT, randomized controlled trial.

The first-in-human trial of dasatinib–quercetin demonstrated safety in a clinical population of idiopathic pulmonary fibrosis⁸³, as did the first phase I trial in older adults with mild cognitive impairment (MCI) or early AD³³. The MCI/early AD open-label study enrolled five participants to assess the safety of an intermittent dasatinib–quercetin treatment protocol and determine BBB penetrance of the compounds. Dasatinib and quercetin were administered orally for two consecutive days followed by a 2-week break, repeated for six cycles. Drug levels were measured in plasma and cerebrospinal fluid (CSF) under fasting conditions at baseline, before drug administration and 80–150 min after the final dose of study drug. After treatment, both dasatinib and quercetin were detectable in the plasma of all participants at higher levels than baseline. Dasatinib was also detected in the CSF in 80% of the participants after treatment; however, quercetin was not. The lack of quercetin detection could be attributed to its low bioavailability, levels being below detection sensitivity, the timing of CSF collection not capturing quercetin owing to its short half-life, or quercetin not penetrating the BBB. Two additional open-label phase I trials testing dasatinib–quercetin in AD (NCT04785300 and NCT05422885) will be important for providing confirmatory safety data, and the ongoing phase II SToMP-AD study (NCT04685590 (ref. 84)) will help to establish efficacy and timing of senescent cell re-emergence (Table 3). An interventional trial has also been initiated using dasatinib–quercetin in treatment-resistant depression, given the observed association between SASP and psychiatric symptoms^55,56. Collectively, the data generated in these early trials are crucial for informing safety, dosing strategies and outcome measures that will guide future trials⁸⁵.

Other nonspecific PI3K-inhibiting flavonoids have been investigated for their senolytic properties. Luteolin and curcumin have shown weak senolytic activity at a dose at which quercetin was ineffective⁸⁶. Fisetin, a natural compound structurally related to quercetin, is found in various fruits, vegetables and teas, has gained considerable attention in the area of senolytic therapies^87,88. Fisetin exhibits a diverse pharmacological profile including antioxidant, anti-inflammatory, antimicrobial as well as neuroprotective properties^87–89. In addition to PI3K, fisetin acts on multiple signalling pathways that have pivotal roles in the senescence process, including BCL-2, PI3K–AKT, p53 and NF-κB. This multifaceted action has led to the recognition of fisetin as a senolytic compound, a discovery that emerged from systematic screening studies involving flavonoid polyphenols and senescent human and murine fibroblasts^86,90. Notably, fisetin has demonstrated the ability to restore tissue homeostasis, mitigate age-related pathologies and extend lifespan in mice⁸⁶. In relevance to neurodegeneration, fisetin has been shown to modulate inflammatory pathways and preserve cognitive function in AD transgenic mouse models⁹¹. It also enhances memory and induces hippocampal long-term potentiation in rats, suggesting its positive action in the brain and ability to ameliorate AD-related symptoms in vivo⁹². Fisetin is currently undergoing clinical testing in approximately 20 clinical trials targeting a range of age-related diseases, including MCI (NCT02741804)⁹³.

Of note, like quercetin, fisetin has low oral bioavailability owing to factors such as low aqueous solubility, high lipophilicity and high first-pass metabolism⁹⁴. To improve the bioavailability of these drugs, strategies may include careful modification of their chemical structures to enhance their pharmacokinetic profiles while maintaining their efficacy⁹⁵, or using novel drug delivery systems^96,97.

HSP90 inhibition.

The essential family of heat shock protein 90 (HSP90) is involved in multiple cellular processes and regulatory pathways such as apoptosis, cell cycle control, cell viability, protein folding and degradation, and various signalling cascades⁹⁸. Dysregulation of HSP90 is implicated in diverse pathologies, including cancer and neurodegenerative diseases such as Parkinson disease⁹⁹ and AD¹⁰⁰. In a senescence assay-based screen of compounds that regulate autophagy, inhibitors of the HSP90 chaperone were identified to have senolytic activity in mouse and human cells, including fibroblasts, mesenchymal stem cells and umbilical vein endothelial cells¹⁰¹. Various HSP90 inhibitors showed dose-dependent senolytic potential, and inhibition of HSP90 delayed the onset of age-related symptoms, extended health span and reduced p16^INK4a levels in a mouse model of a human progeroid (premature ageing) syndrome¹⁰¹.

Na⁺K⁺-ATPase inhibitors.

The Na⁺K⁺-ATPase pump is a crucial transmembrane protein that actively transports sodium and potassium ions against their respective concentration gradients to maintain the cellular osmotic balance and membrane potential¹⁰². Recent studies have reported that senescent cells possess a marginally depolarized plasma membrane and elevated concentrations of H⁺ compared with control cells¹⁰³. A similar aberrant membrane potential phenotype has been described in senescent neurons^10,37,104.

A high-throughput approach was used to determine the in vitro and in vivo senolytic potential of cardiac glycosides, which inhibit the Na⁺K⁺-ATPase¹⁰³. In vivo treatment with cardiac glycosides induced the clearance of senescent tumour cells and of senescent fibroblasts in a mouse model of lung fibrosis¹⁰³. The authors acknowledge that the therapeutic range for cardiac glycosides such as digoxin is relatively narrow. However, they also emphasize that digoxin has been successfully used for many years, and that monoclonal antibodies capable of reversing potential drug overdoses have been developed¹⁰³. Another cardiac glycoside, ouabain, was shown to selectively eliminate cells undergoing oncogene-induced senescence in a mouse model of liver cancer¹⁰⁵. Both studies suggest that the senolytic effects of cardiac glycosides are due to on-target inhibition of the Na⁺K⁺-ATPase, but further research is required to explore the molecular senolytic properties of these compounds, including for application to the CNS.

BCL-2 protein family inhibitors.

The BCL-2 family has a key role in regulation of apoptosis and has broader functions. These include neuronal activity, autophagy, calcium regulation, mitochondrial dynamics and energy production, as well as various other processes essential for maintenance of the health and functionality of cells¹⁰⁶. The BCL-2 family includes both inhibitors and promoters of cell death, collectively orchestrating the intrinsic apoptosis pathway, with mitochondria having a pivotal role. Senescent cells use an anti-apoptotic programme driven by members of the BCL-2 family, including BCL-2, BCL-X_L and BCL-W, to persist within tissues and contribute to chronic inflammation¹⁰⁷.

Inhibition of the BCL-2 pathway can induce apoptosis in some, but not all, senescent cell types^61,107. This discovery led to the identification of BCL-2 family inhibitors as senolytic compounds, including ABT-737, ABT-263 (navitoclax), A-1331852 and A-1155463. ABT-737, an inhibitor of BCL-W, BCL-X_L and BCL-2, was the first in this class shown to specifically induce apoptosis in senescent cells through a BCL family-dependent mechanism¹⁰⁷. In mice, ABT-737 effectively removed senescent cells in the lungs and epidermis, resulting in increased proliferation of hair-follicle stem cells¹⁰⁷. However, ABT-737 has limitations such as poor oral bioavailability and low aqueous solubility, leading to the development of an orally bioavailable derivative, ABT-263, as a pan-BCL inhibitor¹⁰⁸. In addition to its in vitro senolytic activity, ABT-263 has been shown to rejuvenate aged haematopoietic stem cells and effectively clear senescent cells in irradiated or aged mice¹⁰⁹. Nonetheless, notable side effects, including thrombocytopenia and transient thrombocytopathy, have been reported for ABT-263 (ref. 110).

Inhibiting BCL-2 family members can pose challenges in diseases of the CNS. Treatment with ABT-737 has been found to be toxic to stem cell-derived dopaminergic neurons in vitro¹⁰, likely due to the reliance of neuronal development and survival on BCL-X_L activity¹¹¹. ABT-263 does not easily penetrate the BBB, hindering its ability to reach the brain parenchyma¹¹². Nevertheless, studies in tau transgenic mice indicate that ABT-263 reduces senescent cells in the brain and improves behaviour⁵⁷. Despite their demonstrated senolytic effects, the clinical translation of BCL-2 family inhibitors to CNS disorders is collectively constrained by potential side effects, off-target effects on platelets and limited CNS penetrance. Although systemic removal of senescent cells might positively impact the brain regardless of BBB penetrance (discussed below), future research efforts are still needed to develop more specific, safe and brain-penetrant BCL-2 family inhibitors for use as CNS senolytic therapies.

More recently, an ABT-263-based BCL-X_L proteolysis-targeting chimera (PROTAC) senolytic has been developed to target BCL-X_L to the cereblon E3 ligase for degradation¹¹³. Additionally, to selectively target senescent cells, researchers leveraged the elevated activity of the lysosomal β-galactosidase that is common in senescent cells, developing a drug delivery system based on encapsulating drugs with galacto-oligosaccharides^114,115. In a mouse model involving chemotherapy-induced senescence, gal-encapsulated cytotoxic drugs effectively targeted palbociclib-induced senescent tumour cells, leading to improved regression of tumour xenografts. In another mouse model of pulmonary fibrosis, gal-encapsulated cytotoxic drugs or ABT-263 specifically targeted senescent cells, resulting in reduced collagen deposition and restored pulmonary function. Notably, the gal-encapsulation approach mitigated the toxic side effects associated with senolytic drugs¹¹⁴. Furthermore, Nav-Gal, engineered by linking the BCL-2 family inhibitor ABT-263 to an acetylated galactose group, selectively eliminated senescent cells both in vitro and in vivo, demonstrating reduced platelet toxicity and ameliorating the common issue of thrombocytopenia associated with ABT-263 treatment¹¹⁶. Although this approach is promising, targeting β-galactosidase for CNS conditions may not be appropriate as many non-senescent neuronal cell populations express high levels of β-galactosidase activity^9,117 (Fig. 2), as discussed further below.

USP7, MDM2 and p53–FOXO4 interaction inhibitors.

Modulators of the p53 pathway have gained attention as potential senolytic agents owing to the role of p53 in controlling senescence and apoptosis, particularly through its regulation of BCL-2 family proteins^118,119. One note-worthy example is forkhead box protein O4 (FOXO4), an anti-apoptotic transcription factor that is upregulated in some senescent cell types. FOXO4 prevents cell death by binding to and sequestering p53 within the nucleus¹²⁰. To target this interaction, a senolytic peptide named FOXO4-d-Retro-Inverso (FOXO4-DRI) was developed. FOXO4-DRI disrupts the binding between FOXO4 and p53, which allows p53 to translocate to the cytosol and induce apoptosis in senescent cells. FOXO4-DRI has demonstrated selective removal of senescent cells and has shown promise in alleviating age-related symptoms across organ systems in animal models¹²⁰.

p53 protein levels are tightly regulated through ubiquitylation by the E3 ubiquitin ligase murine double minute 2 (MDM2). Inhibitors that disrupt the MDM2–p53 interaction, such as UBX0101 and RG7112 (RO5045337), as well as inhibitors of the de-ubiquitinating enzyme ubiquitin-specific peptidase 7 (USP7), serve to increase and stabilize p53 levels. RG7112, UBX0101 and USP7 inhibitors, such as P5091 and P22077, cleared various peripheral senescent cells by upregulating p53 (ref. 121).

Moving into clinical applications, however, has faced challenges. Although UBX0101 has been shown to induce apoptosis of senescent synoviocytes, and a phase I study (NCT03513016) suggested that it has clinically meaningful effects on pain and function in patients with knee osteoarthritis, preliminary reports from the follow-up phase II trial of UBX0101 for osteoarthritis suggest that it has not demonstrated significant efficacy compared with placebo¹²². The authors speculated that possible reasons for this could include the dosing route, patient and/or investigator expectedness, and gender dimorphism in pain reporting. Manipulating p53 levels in senescence regulation requires careful consideration, as p53 can have both promoting and inhibitory effects depending on cell type and stress levels. Adding complexity to the therapeutic approach, p53 activity can induce p21-dependent senescence. Additionally, some p53-targeting agents are peptide based, which may limit their applicability to address brain senescence. As mentioned above, senescent cells have been reported in Parkinson disease post-mortem human brain tissue and model organisms^10,12,47,123. Elevated p53 levels and activity were observed in the brains of patients with Parkinson disease, as well as in Parkinson disease animal and cellular models, mostly correlating with DNA damage¹²⁴. p53-dependent brain senescence has also been reported in Huntington disease and AD¹²⁴, suggesting that p53 may be an especially appealing target for CNS senescence. Recently, it was reported that a ketogenic diet induced p53-dependent cellular senescence in multiple tissues, including the brain¹²⁵. These findings may have serious implications for humans on the ketogenic diet; however, we note that immunoblotting with an antibody to β-galactosidase was the primary readout, which may reflect changes in lysosomal function rather than senescence. Further research is needed to unravel the complicated molecular mechanisms by which p53 regulates senescence, with potential implications for clinical applications.

HDAC inhibitors.

Histone deacetylases (HDACs) modulate gene expression by removing acetyl groups at the lysine residues of histones and non-histone proteins. They impact many cellular functions including differentiation, proliferation and survival¹²⁶. Senescent cells show changes in H3 acetylation and BCL-X_L expression^107,127,128. In vitro, the FDA-approved HDAC inhibitor panobinostat cleared cancer cells with chemotherapy-induced senescence. This activity was associated with decreased BCL-X_L expression and increased H3 acetylation¹²⁹.

BRD4 inhibitors.

Bromodomain and extra-terminal domain (BET) family proteins act as transcriptional regulators¹³⁰. The family member BRD4 is well-characterized for its role in transcriptional elongation¹³¹. In a high-throughput screen of compounds alongside bio-functional analysis, the BET family protein degrader (BETd) was identified as senolytic¹³². The authors showed that inhibition of the BET family protein BRD4, either with chemical inhibitors or gene-specific RNA interference, induced senolysis. The hetero-bifunctional PROTAC ARV825 proved particularly effective, as it recruits BET family proteins to the E3 ubiquitin ligase cereblon, leading to their degradation¹³³. The senolytic effects of ARV825 were mediated by attenuating non-homologous end joining (NHEJ) and enhancing autophagic gene expression. Specifically, the authors noted that ARV825 induces senolysis by exacerbating DNA double-strand breaks through targeting at least two independent mechanisms of the NHEJ machinery in senescent cells: first, inhibition of gene expression of XRCC4, which encodes a protein required for NHEJ repair; and secondly, blockade of the recruitment of 53BP1, an adaptor protein necessary for the assembly and activation of the DNA repair machinery, to double-strand break sites¹³². Treatment with ARV825 was sufficient to eliminate senescent cells by targeting BRD4, as confirmed by small interfering RNA (siRNA)-based gene-specific experiments, which showed that ARV825 reduced levels of BRD3 and BRD4, but not BRD2. Furthermore, the compound showed senolytic activity in the livers of obese mice and cleared senescent hepatic stellate cells, as well as chemotherapy-induced senescent cells in mice¹³².

OXR1 inhibitors.

Oxidation resistance gene 1 (OXR1) acts as a sensor and regulator of cellular oxidative stress. It is involved in the transcriptional networks needed to detoxify reactive oxygen species and modulate cell cycle and apoptosis¹³⁴ and is upregulated in senescent human fibroblasts¹³⁵. Interestingly, OXR1 is the target of the previously identified senolytic piperlongumine, a natural compound found in long pepper^135,136. Piperlongumine was identified by screening a library of rationally selected compounds for their senolytic activity in human senescent fibroblasts¹³⁶. The initial study reported that piperlongumine kills senescent fibroblasts without the induction of reactive oxygen species; however, it was later demonstrated that piperlongumine binds to OXR1, leading to its degradation via the ubiquitin–proteasome system specifically within senescent cells. The reduction of OXR1 induces apoptosis exclusively in senescent cells, likely by increasing their vulnerability to oxidative stress¹³⁵. In a recent computational study investigating natural senotherapeutics to find candidates that could replace dasatinib, based on their similarity in gene expression effects, piperlongumine was identified as the highest-potential substitute, suggesting that further development of the compound is warranted¹³⁷.

PUMA/NOXA activators.

PUMA and NOXA are pro-apoptotic genes that are functionally repressed by BCL-2 family molecules in senescent cells¹³⁸. Through screening natural products, procyanidin C1 (PCC1), a polyphenolic component of grape seed extract, was identified as being capable of ameliorating the SASP at low concentrations and exhibiting senolytic effects at higher concentrations¹³⁹. It was hypothesized that PCC1-induced apoptosis in senescent cells is partially mediated by NOXA and PUMA, accompanied by increased reactive oxygen species production and mitochondrial dysfunction¹³⁹. Additionally, PCC1 ameliorated physical dysfunction and prolonged survival in mice.

Senomorphics

Senomorphics, also known as senostatics, mitigate the toxicity of senescent cells without removing them from the tissue, with the ultimate goal of retaining the cells’ initial physiological functions (Table 2 and Fig. 3). This section focuses on the molecular pathways targeted in senomorphic drug development.

ATM inhibitors.

The ataxia–telangiectasia mutated (ATM) kinase is crucial for DNA damage response signalling and is known to activate NF-κB in response to stress. NF-κB is a transcription factor involved in the regulation of the immune response, inflammation, cell survival, cellular senescence and organismal ageing¹⁴⁰. ATM kinase mediates the equilibrium between senescence and apoptosis. Active ATM promotes autophagy, with a specific emphasis on sustaining the lysosome–mitochondrion connection, boosting senescence while restraining apoptosis¹⁴¹. Disruptions in autophagy mechanisms, however, have been demonstrated to elevate DNA damage levels, thereby promoting the development of cancer and neurodegenerative diseases¹⁴². Activated ATM has been reported in senescent cells, and genetic and pharmacological (using KU-55933) inhibition of ATM resulted in a reduction in NF-κB activity, markers of senescence and the SASP in cells as well as in mouse models of ageing¹⁴³. This treatment also extended the health span of progeroid mice, indicating that ameliorating ATM-mediated NF-κB activity can slow the progression of ageing¹⁴³.

mTOR inhibitors.

The PI3K-related kinase mechanistic target of rapamycin (mTOR) regulates numerous cellular pathways including apoptosis, growth and autophagy^144,145. mTOR inhibition has emerged as an avenue to improve longevity and ameliorate the impact of cellular senescence. The mTOR inhibitor rapamycin, an FDA-approved immunosuppressive drug, has repeatedly proved its ability to extend lifespan in various organisms, ranging from yeast to mice^146,147. Rapamycin promotes autophagy, a cellular process that has a crucial role in recycling damaged components and maintaining cellular health. Although not a senolytic, rapamycin has gained attention for its potential to suppress the SASP¹⁴⁸. Although rapamycin holds promise for extending lifespan and curbing the SASP, clinical applications have encountered challenges, including metabolic dysregulation; impaired wound healing, a process in which senescent cells appear to have a beneficial role¹⁴⁹; and the development of hyperlipidaemia¹⁵⁰. Despite these challenges, the positive outcomes observed in various animal models of ageing and neurodegeneration have propelled rapamycin into clinical trials to assess its potential in treatment of AD (NCT04200911, NCT04629495), although it is not known whether senescence-associated outcomes will be assessed.

NF-κB inhibition and SASP modulation.

The transcription factor NF-κB has a dual role, acting as both a guardian against and a potential contributor to neuroinflammation and the accumulation of senescent cells in the brain¹⁵¹. It mediates neuroinflammation by enhancing the expression of pro-inflammatory molecules, many of which are part of the SASP, such as inducible NO synthase (iNOS), cyclooxygenase 2, TNF and interleukins (for example, IL-1 and IL-6). Under normal conditions, the activity of NF-κB is restrained by inhibitors known as IκB proteins. However, in response to stress or inflammation, IκB inhibitors are phosphorylated by the IκB kinase (IKK) complex, unleashing NF-κB and triggering a cascade of immune responses¹⁵². Studies have shown that enhanced NF-κB activity can lead to neuroinflammation and an increased burden of senescent cells in the brain¹⁵¹. Conversely, inhibition of the binding between NF-κB and IKK, using a peptide inhibitor, ameliorated senescence in progeroid mice¹⁵³. Building on this, researchers developed a small molecule, SR12343, which disrupts the interaction between NF-κB and IKK. This molecule significantly reduced the release of SASP factors, alleviated senescence and improved markers of senescence in mice¹⁵⁴. Importantly, SR12343 can traverse the BBB, making it a compelling candidate for evaluation in the context of neurodegenerative conditions¹⁵⁵. In a recent study using SR12343, it was found to lower markers of cellular senescence and inflammation in the liver, skeletal muscle and blood, which were elevated in response to chemotherapy. Additionally, it was reported that SR12343 effectively counteracted chemotherapy-induced skeletal muscle wasting and dysfunction¹⁵⁶.

The p38 pathways regulates the transcriptional activity of NF-κB¹⁵⁷. Consequently, inhibiting the p38 pathway has emerged as a promising avenue to address the detrimental effects of the SASP and reduce age-related inflammation, particularly in the context of neurodegenerative conditions. The p38 inhibitor SB203580 has been shown to reduce the expression of SASP factors in human senescent cells¹⁵⁸. Similarly, next-generation p38 inhibitors, such as UR-13756 and BIRB 796, have demonstrated promising results in limiting the expression of IL-6. By reducing the p38-dependent transcriptional activity of NF-κB on pro-inflammatory genes, these inhibitors mitigate SASP-related effects, marking them as potential therapeutic candidates for age-related inflammatory conditions¹⁵⁹.

The regulation of the SASP is complex. Recent findings indicate that the immune gene transcriptional effector C/EBPβ cooperates with NF-κB to activate SASP genes¹⁶⁰. JAK–STAT signalling is involved in the regulation of C/EBPβ and vice versa. Inhibition of the JAK–STAT signalling pathway by ruxolitinib suppresses the transcriptional activity of C/EBPβ. This suppression led to a reduction in systemic inflammation, attributed to SASP repression, and an improvement in the overall fitness of aged (24-month-old) mice¹⁶¹. These findings underscore the potential of modulating stress kinase p38 as a senomorphic approach to combat neuroinflammation, a common feature of neurodegenerative disorders. To tackle neuroinflammation in neurodegenerative diseases, a brain-permeable and orally available p38 inhibitor, MW150 (also known as MW01-18-150SRM), is currently in phase II trials for AD (NCT05194163).

Although strategies to reduce NF-κB activity hold promise in the realm of senomorphic therapies, inhibition may lead to side effects owing to NF-κB involvement in diverse biological processes. An alternative approach to reducing inflammation, perhaps with a better safety profile, involves the use of the nonsteroidal anti-inflammatory drug, ibuprofen. This strategy has been shown to mitigate senescent cell accumulation, neuroinflammation and cognitive dysfunction in mice¹⁵¹. However, clinical trials using ibuprofen as monotherapy for AD have not significantly impacted disease progression or outcomes^162,163, but perhaps could be a useful combination therapy adjunct.

Harnessing the immune system

A healthy immune system actively clears senescent cells and prevents their accumulation. Therapeutic approaches to boost senescent cell clearance by the immune system are under investigation. These include enhancing the recognition of senescent cells by the immune system, promoting immune cell activity or modulating immune responses to improve the clearance of senescent cells from tissues. Generally, modulating the immune system for senolysis represents a promising approach.

The body’s natural defence against senescent cells involves immune surveillance mechanisms, with immune cells such as macrophages, T cells, natural killer cells (NK cells)^164,165 and microglia in the brain^44,166 tasked with removing these cells. However, age-related immunosenescence and the ability of senescent cells to evade immune detection – for example, by expressing immunosuppressive molecules such as PDL1 and PDL2, by producing tolerogenic major histocompatibility complex class I variants and by secreting factors that attract immunosuppressive cells¹⁶⁷ – can lead to their accumulation in tissues¹⁶⁸.

Notably, some senescent cells upregulate specific surface markers that are being explored as therapeutic targets. Urokinase-type plasminogen activator receptor (uPAR) has been shown to be upregulated in several senescent cells, including mouse lung adenocarcinoma cells induced to senesce by MEK and CDK4/6 inhibition, oncogene-induced senescence in mouse hepatocytes and culture-induced senescence in mouse hepatic stellate cells¹⁶⁹. Mass spectrometry analyses comparing proliferating human fibroblasts with those undergoing replicative senescence revealed an increase in the surface molecule dipeptidyl peptidase 4 (DPP4; also known as CD26) in senescent cells¹⁷⁰. Transcriptome data from senescent vascular endothelial cells indicated an enrichment of the transmembrane glycoprotein nonmetastatic melanoma protein B (GPNMB)¹⁷¹. Collectively, uPAR, DPP4 and GPNMB were identified as ‘seno-antigens’ and targeted for immune-based interventions as described below.

To bolster immune clearance of senescent cells, chimeric antigen receptor-T (CAR-T) cells have been used. In this approach, T cells from patients are genetically engineered to target and eliminate cells expressing specific antigens. Utilizing uPAR as a target antigen, CAR-T cells have been shown to effectively clear senescent cells, mitigating various pathologies in mouse models of senescent lung adenocarcinoma and liver fibrosis¹⁶⁹. Similarly, antibody-based therapies targeting the senescent cell surface protein DPP4 have demonstrated efficacy in eliminating senescent cells in vitro¹⁷⁰.

A vaccine-like protocol aimed at senolysis has emerged as another possible approach to mitigate senescent cell burden. Immunizing mice against GPNMB, which is elevated in senescent vascular endothelial cells and in samples from patients with atherosclerosis, reduced GPNMB-positive cells, improved metabolic parameters in mice fed a high-fat diet, reduced atherosclerotic plaque burden in Apoe-knockout mice, improved age-related phenotypes and ultimately extended the lifespan of progeroid mice¹⁷¹. Developing immune- and vaccine-based approaches to eliminate senescent cells is intriguing, but identifying senescence-specific antigens is crucial to avoid potential autoimmunity against non-senescent cells. To date, none of the senescence-associated surface markers seems truly exclusive for senescent cells, as GPNMB is expressed across tissue and cell types and has diverse roles in inflammation and disease^172,173; DPP4 is known to be ubiquitously expressed and involved in immunoregulatory actions¹⁷⁴; and uPAR is present on many immune cells and is detected in various bodily fluids including plasma and seminal fluid, and in extracellular matrix¹⁷⁵.

In addition to the challenge of identifying markers exclusively expressed on senescent cells, targeting CNS senescence through immune modulation faces other challenges. Immune cell trafficking into the CNS is tightly regulated by the BBB, permitting entry only for specific immune cell subsets¹⁷⁶. Additionally, immune-based approaches trigger an inflammatory response, which could be problematic in the context of neurodegenerative diseases, where neuroinflammation is present. Therefore, while harnessing the immune system holds great potential for senescent cell clearance, the impact of on-target, off-tissue effects and unintended immune responses must be carefully evaluated to minimize potential harm.

Future perspective

Many ‘first-generation’ senolytic drugs such as quercetin, fisetin and piperlongumine, lack a single precise target and instead affect multiple pathways that are implicated in cellular ageing. These broad-spectrum senolytics have the capacity to impact multiple senescence pathways simultaneously¹⁷⁷; however, they present a challenge in predicting efficacy owing to the considerable heterogeneity between senescent cell populations. Drug discovery programmes based on identifying molecules that are exclusively upregulated in senescent cells offer opportunities for more precisely targeted senolytic therapeutics. Other exciting senotherapeutic approaches, currently in preclinical stages, include cellular reprogramming¹⁷⁸; targeting mitochondrial function^179–181; targeting the cGAS–STING pathway^182,183; and various genome-targeting approaches including G-quadruplex structures¹⁸⁴, DNA methylation¹⁸⁵ and retrotransposon activity¹⁸⁶. These diverse and innovative approaches reflect a growing enthusiasm for senotherapeutics and underscore the need for continued research to fully realize their potential for CNS conditions.

Challenges in developing CNS senotherapies

Senotherapeutic drug development faces several challenges, some of which are common to CNS drug development more broadly and contribute to the long development timelines of such therapeutics¹⁸⁷. Here, we consider challenges that are particularly important in the development of CNS senotherapeutic drugs and discuss strategies to address them.

Senescence-specific targeting

The need to achieve selectivity for therapeutic targets and specific cells, while sparing healthy cells, to achieve a favourable safety profile is a common challenge for drug discovery. Senotherapeutics that impact cellular mechanisms that are vital for neuronal viability and function, such as Na⁺K⁺-ATPase inhibitors¹⁸⁸, BCL-2 inhibitors¹⁸⁹ and NF-κB inhibitors¹⁹⁰, pose a particularly substantial risk. Understanding cellular senescence in various disease states and identifying relevant cell types are crucial to identify and select optimal senotherapeutic targets¹⁹¹. Drug development, and most senescent cell identification, occurs in laboratory models. Discerning the differences between human disease and experimental models is a crucial step in safely and effectively translating senotherapies to the clinic. We anticipate that technological advances in single-cell and spatial multi-omics, along with advanced computational and AI approaches capable of analysing the vast datasets, will enable the identification of tractable and specific targets for senotherapeutic interventions²⁸.

New drug modalities in development could help to increase specificity for therapeutic targeting. For example, PROTACs, as mentioned earlier, are an innovative mechanism that capitalizes on the cell’s ubiquitin–proteasome system¹⁹². They hold promise for specificity, reduced side effects and the ability to hit targets previously deemed ‘undruggable’. However, their size presents a significant hurdle to optimization of CNS delivery, necessitating research efforts to improve their design for such applications. Beyond modalities that directly engage the target proteins, others such as gene therapy and RNA-based therapies offer versatile approaches for gene silencing or functional restoration and are in early stages of development¹⁹³.

BBB penetrance

Favourable properties of molecules that target the CNS include small size, hydrophobicity optimal for penetrating the BBB without becoming trapped in the lipid bilayer or losing blood solubility, and low affinity for the P-glycoprotein transporter¹⁹⁴. Peptides and antibodies offer higher specificity compared with small molecules, but they have low BBB penetrance and face challenges related to stability, solubility and incompatibility with oral administration. Advances in engineering antibodies and designing peptides that are more resistant to degradation, along with new delivery systems and BBB-crossing strategies, such as tagging with molecules that allow uptake via receptor-mediated transcytosis (for example, TfR¹⁹⁵ and CD98hc¹⁹⁶) are making them more viable for CNS applications¹⁹⁷, although this is a complex and ongoing challenge. Even when drugs penetrate the brain, determining the correct dosage is challenging because drug metabolism in the brain can differ from that in other organs^198,199.

Notably, the cells that comprise the BBB are susceptible to becoming senescent. Vascular smooth muscle cells were identified as the most prominent senescent cell type across 50 healthy human tissues³¹. A separate study identified a significant increase in senescence gene expression in vascular cells isolated from post-mortem human brains with advanced Braak stages of AD compared with control cases²⁰⁰. These findings suggest that senotherapies may benefit the brain by targeting vascular cells, potentially without the need for BBB penetration or direct brain exposure. Although BBB penetrance is a key consideration for CNS drug development, there is some evidence that clearing senescent cells in peripheral tissues may benefit brain function. For example, ABT-263, which does not easily penetrate the BBB, has shown benefits to the CNS following whole-body irradiation²⁰¹ in mouse models that develop AD neuropathology^57,202 and in glioblastoma models^203,204. Although the senescence-inducing insults (for example, radiation, toxic tau and amyloid-β, or cancer, respectively) may have allowed ABT-263 to penetrate the BBB, it is also possible that the benefits were an indirect result of clearance of senescent peripheral cells, which have been shown to impact senescent cell accumulation in the brain²⁰⁵. Thus, senolytic therapies may improve brain function by clearing senescent immune cells and cells of brain vasculature without a need to fully penetrate the parenchyma.

Although the ‘hit-and-run’ strategy of intermittent dosing has been used to minimize potential adverse events owing to prolonged systemic levels of senolytics, an alternative approach is local delivery. This strategy was used in a recent open-label phase I trial for UBX1325 (foselutoclax), a senolytic small-molecule inhibitor of BCL-X_L, in patients with advanced diabetic macular oedema (DME)²⁰⁶. In the study, UBX1325 was administered via intravitreal injection, keeping plasma levels of UBX1325 and its active parent molecule below the lower limit of quantification and minimizing adverse events linked to systemic BCL-X_L inhibition. Visual acuity improvements were observed in six of eight patients at 12 weeks, and in five of eight patients at 24 weeks. The favourable safety profile supports the ongoing phase IIb study efforts (Table 3).

Risks of neuronal senolysis

The question of whether postmitotic neurons that have entered a senescence-like state should be eliminated requires careful consideration. The therapeutic elimination of neurons is not unprecedented, with the best-known example of this being the surgical removal of small volumes of brain tissue to treat drug-resistant epilepsy²⁰⁷. Evidence to support the clearance of senescent neurons includes preclinical studies in tauopathy mice that demonstrate benefits from the clearance of NFT-bearing neurons⁹. Neurons with NFTs maintain the ability to respond to inputs and remain active in vivo²⁰⁸ but contribute to pathogenic tau spread in an activity-dependent manner²⁰⁹. Similarly, senescent neurons maintain electrical activity, but with altered function. For example, inducing a senescence phenotype in excitatory neurons by deleting Gdf11 causes neuronal hyperexcitability with reduced synaptic inputs in vivo¹⁰⁴. A similar hyperexcitable phenotype with decreased synaptic protein expression was observed in an in vitro model of neuronal senescence generated by transfecting iNeurons with CDKN2A²⁰⁸. Removing SATB1 in human dopaminergic neurons drives a senescence-like phenotype in vitro, whereby the cells maintain spontaneous pacemaking activity but show significant differences in maintenance of response to positive current injections¹⁰. Regulated neuronal activity is crucial for proper brain function, warranting a better understanding of the consequences of altered excitability in senescent neurons.

Senescent cells induce paracrine senescence in neighbouring cells through their SASP. Neuronal SASP may contain toxic aggregate-prone proteins⁴⁷, which could facilitate senescence spread between distant regions particularly via well-connected, anatomically linked neurons²¹⁰. A link between intraneuronal tau^19,98 and α-synuclein deposition¹²³ and senescence has been established. Given that neurons increase the release of α-synuclein after becoming senescent⁴⁷, it is tempting to speculate that neuronal senescence is a key process mediating the transfer of aggregate-prone proteins between neurons and across circuits. Senescent neurons also spread senescence to non-neuronal cells including to astrocytes through traditional SASP mechanisms²⁰⁸ and to microglia through phagoptosis^44,166. Specifically, neurons that contain NFTs expose phosphatidylserine on their cell membranes, which signals microglial phagocytosis¹⁶⁶. Upon engulfing the NFT-bearing neuron, the microglia become senescent, resulting in their decreased tissue surveillance and their release of partially digested NFT toxic fragments into their environment, which amplifies pathogenesis⁴⁴ (Fig. 1). In summary, the connectivity of senescent neurons within complex neuronal circuits and their close interaction with glia cells may allow them to propagate neuropathology to multiple brain cell types, which provides rationale in favour of their clearance. However, in some neurodegenerative models associated with aberrant microglial activity, positive outcomes have been seen from retaining damaged neurons by interrupting phagoptosis^211–214. These studies highlight the nuanced, and important, choice of senotherapy approach (senolytic versus senomorphic) depending on context and condition, especially when targeting cells and tissues with limited regenerative capacity.

Measuring CNS target engagement

Target engagement, which measures the interaction of a compound with its intended molecular target within a living system, provides insights into the drug’s pharmacological profile, mechanism of action and potential efficacy. Evaluation of precise target engagement in preclinical senolytic studies often involves histological analysis, as well as protein and gene expression measurements on mouse brain tissue^9,45,57. These methods offer detailed insights into the cellular and tissue distribution of senolytic effects. However, assessment of target engagement in living humans requires in vivo imaging strategies or surrogate measures in accessible biofluids.

MRI provides information on the effects of a drug on brain morphology and physiology²¹⁵ and has been used to assess the impact of senolytics in a tauopathy mouse model⁹. Although MRI cannot provide direct evidence for the action of senotherapeutics on pro-survival or SASP pathways within senescent cells, it allows for the longitudinal tracking of disease progression in the brain, providing a surrogate marker of drug effects in the target tissue. Positron emission tomography (PET) tracks the accumulation of radiolabelled drugs in the body²¹⁶, with ongoing development of senescence-specific PET tracers focusing on labelling lipofuscin²¹⁷ or β-galactosidase²¹⁸, which appear when mitotically competent cells enter senescence. However, these tracers may have limited applicability to cerebral tissues as lipofuscin deposition and elevated β-galactosidase activity can also be detected in healthy CNS tissue throughout the lifespan^9,117,219 (Fig. 2).

Other strategies for identifying surrogate markers of target engagement include characterization of the surface proteome of senescent cells to discover seno-antigen-specific proteins unique to these cell populations^220,221, which can then be visualized using clinical imaging techniques, such as immuno-PET²²². A recent report suggests that senescent cells produce unique oxylipins that are released upon their death, raising the possibility of developing senolysis-specific biomarkers²²³. Although the identification of PET tracers specific to brain senescent cells could represent a significant advance, it remains unclear whether the small percentage (2%) of sporadically dispersed senescent cells¹⁹ would produce a strong enough signal to be detectable in PET imaging.

The limitations with imaging senescent cells in the brain have increased the reliance on biofluid analytes as surrogate markers of target engagement. Mouse studies suggest that senescent T cell levels in blood could serve as a potential surrogate for senescent burden in solid tissues, including the brain²⁰⁵. In this regard, p16^INK4a (p16) mRNA expression in peripheral blood T lymphocytes has emerged as a quantifiable measure of total senescent cell load²²⁴ and is being used in clinical trials^62,225,226, including an ongoing phase II trial for AD⁸⁴. Other commonly used surrogate markers for senescent cells include circulating SASP factors²²⁷ such as cytokines, chemokines, proteases and growth factors^82,83. Similar to measuring a SASP index, as mentioned earlier, the first phase I clinical trial to test senolytics in older adults with MCI or early AD, used a SASP cytokine panel to evaluate senescence-associated markers of inflammation in the blood and CSF³³. However, the challenge with using SASP factors is that they can be secreted for reasons other than cellular senescence, complicating the interpretation of outcomes. Ongoing efforts to understand how the secreted milieu of senescent cells varies by cell type or context will help to clarify the utility of these promising biomarkers^228,229. Moreover, results from placebo-controlled trials, such as the phase II SToMP-AD study, are needed to interpret how SASP readouts relate to senolytic effects⁸⁴.

Methodological advances with omics technologies are paving the way for a deeper understanding of molecular changes associated with CNS diseases²³⁰. These approaches may allow for the identification of rare and distinct senescent cell populations and omics changes in response to senolytic therapies. For example, whereas traditional SASP protein markers did not change significantly in the first phase I trial for AD³³, lipidomics and transcriptomics analyses revealed analytes that may be more sensitive to senolytic treatment⁸⁵. While we are cautious not to overinterpret the results from the open-label trial, especially given the small sample size, the approach underscores the ongoing efforts to identify biomarkers specific to senescent cells and their clearance.

Applying AI to senotherapy development

The complex and heterogeneous phenotypes of senescent cells requires innovative approaches in developing effective senotherapeutics. Advances in experimental platforms and omics technologies, coupled with AI and ML, enable the extraction of relevant information for chemical activity and functional genomics in drug discovery²³¹. AI refers to the simulation of human intelligence in machines for performing tasks, making decisions and adapting by learning from their experiences. ML, a specific AI technique, involves training computer algorithms to learn from data and take actions based on the learned knowledge²³¹. Predictive modelling using ML algorithms expedites decision-making²³², identifies potential drug candidates with a higher likelihood of success, and reduces costs and development time^233,234.

In this section, we briefly review AI and ML applications in key steps of senotherapeutic development, including identification of senescent cells, identification of senescence targets, assessment of druggability and virtual screening compounds, and highlight major challenges in these areas. Overall, the application of AI and ML in senescence research is still in its early stages but holds potential to accelerate drug development in this area.

Identifying senescent cells

Senotherapeutic drug development requires accurate identification of senescent cells. ML approaches can be used to classify and characterize senescent cells using morphological and omics (transcriptomics, proteomics and methylomics) data²³⁵. For instance, ML methods such as multivariate regression analysis^236,237 and k-means clustering²³⁸ have been used to identify senescent cells using bulk transcriptomic data. In multivariate regression analysis, the ‘known’ senescent genes were treated as variates. For single-cell RNA-sequencing data, Teo et al.³² used pseudotime trajectory analysis to identify senescent cell subpopulations. This is an especially powerful approach as it captures cells and cell functions at various stages of senescence progression. Traditional methodologies focus primarily on the end point senescent state at which lysosomal function and cytoskeletal abnormalities are enriched. However, trajectory analyses suggest that cellular respiration and active transportation are transiently elevated in the middle trajectories, which could be interesting targets to explore as their modulation could potentially prevent cells from entering the terminal senescence state.

Others are investigating morphology-based approaches to identify and classify senescent cells. A convolutional neural network (CNN) system, Deep Learning-Based Senescence Scoring System by Morphology (Deep-SeSMo), was trained on phase-contrast images of human umbilical vein endothelial cells treated with senescence-inducing agents²³⁹. Thousands of single cells were given a senescence probability score, predicted as senescent or control, and then compared with predetermined answers. Weighted, automatic and iterative optimization was used to improve the accuracy of the system, resulting in an accuracy of 0.93 and area under the curve of the receiver operating characteristic of 0.98.

Although these approaches hold great potential, limitations include a lack of rigorous wet-lab experiments to confirm the informatic predictions. For example, studies with cell culture validation data commonly rely on SA-β-gal as a gold-standard marker for senescence, which may be misleading, especially for brain cells.

Predicting novel senescence-associated genes

Intersecting cancer driver genes and ageing-associated genes identified dual-purpose targets for the treatment of cancer and ageing and revealed that senescence pathways were upregulated in one subgroup²⁴⁰. A comprehensive transcriptomic network analysis of human tissue-specific RNA-sequencing data from the Genotype-Tissue Expression Project (GTEx) identified co-expressed gene modules enriched for senescence-associated genes³¹. This approach revealed 51 senescence-associated genes conserved across all human tissues and identified regulators centred around CDKN1A (p21)³¹. Although p21 is not a novel senescence-associated molecule, the study represents a first effort in systematically identifying senescence pathways and regulators across multiple human tissues³¹.

Predicting novel senotherapeutic compounds

Research teams are also using ML approaches to discover novel senolytics^239,241. A dataset with 58 known senolytics and 2,465 non-senolytic drugs, each characterized by 200 molecular and structural properties²⁴¹, was used to train SVM and Random Forest-based classifiers to predict new senolytics. The trained classifiers were then used to screen diverse chemical libraries and identified several potential new senolytic compounds. Three compounds, all plant-derived molecules (oleandrin, periplocin and ginkgetin), showed evidence of senolysis using in vitro models. Mechanistically, oleandrin is a cardiac glycoside that inhibits Na⁺K⁺-ATPase; periplocin, also a glycoside, induces apoptosis through AMPK–mTOR or ERK signalling; while ginkgetin, a naturally occurring biflavonoid, acts on multiple pathways crucial to senescence, including apoptosis induction, cell cycle arrest and JAK–STAT and MAPK signalling.

A separate approach used a morphology-based CNN model, Deep-SeSMo, to calculate senescence scores in a drug screen of 80 compounds tested for their senolytic ability. The readout was percentage of SA-β-gal-positive cells, without presenting data on total cell counts, making it unclear whether senolysis occurred. One possibility is that these experiments may have identified senomorphics that impact lysosomal function and SA-β-gal activity, which is consistent with the senotherapeutic agents chosen to train the system, metformin and nicotinamide mononucleotide, both of which act as senomorphics, not senolytics (Box 1).

These early studies showcase the power of AI and/or ML to rapidly identify multiple potential senotherapeutics, increasing the need for validation in appropriate brain cell culture systems and in vivo models to understand the true predictive potential of these innovative models.

Predicting druggability

Although identifying and selecting a target is a crucial initial step in combatting senescence, the target’s druggability remains a decisive factor in determining the viability of potential therapeutic interventions. Druggability refers to the likelihood that a specific biological target can be effectively modulated by a drug-like molecule^242,243. This aspect is particularly important when exploring potential therapeutic interventions for senescence owing to the complex pathways involved and the diverse nature of cellular senescent responses. For a target to be considered druggable, it should be accessible to the intended drug, whether that is a small molecule or a larger biological agent. The probability of successfully drugging a given target is greatly improved if its structure can be elucidated using methods such as NMR, X-ray crystallography or cryo-electron microscopy¹⁷⁹, as it allows for the use of structure-based drug design. This approach, which uses computer algorithms to identify potential binding pockets in the relevant domains of the target protein and virtually screen them for chemical matter that is likely to engage with these regions before commencing experimental work, greatly cuts down on the number of compounds needed in phenotypic screening approaches relative to the conventional ‘brute-force’ approach, in which large chemical libraries are screened²⁴⁴. Notably, recent advancements in AI systems, such as AlphaFold²⁴⁵ have the potential to revolutionize protein structure prediction by reducing reliance on laborious experimental methods, thereby significantly broadening the scope of available targets for drug design explorations.

Predicting off-target effects

Using systems biology approaches^246,247, including network modelling and integration of multi-omics data, can provide comprehensive insights into the global effects of senescence-targeting drugs beyond the intended target. These tools can facilitate the identification of potential off-target effects, discover novel pathways impacted by the drug or drug combinations and optimize therapeutic efficacy while minimizing adverse effects.

Conclusions

Over the past 5 years, research on cellular senescence in the brain has evolved from phenomenology to clinical trials. Overcoming challenges in CNS drug delivery, deepening our understanding of senescence-associated mechanisms and appropriate targets, and discerning when to use senolytics versus senomorphics are essential to advance these strategies. Similarly, refinement of terminology in the field to reflect specific types of senescent cell (for example, replicative, physiological, stress-induced and postmitotic senescent), will enable more precise drug development as each may require different treatment strategies including choice of senotherapeutic, route of administration, dose and scheduling protocol.

Developing dementia has become the most feared health concern among adults worldwide^248–250. As the global population continues to live longer, the prevalence of dementia is expected to rise significantly, creating an urgent need for effective therapeutic strategies. Interdisciplinary approaches, AI and ML advancements and a comprehensive view of CNS senotherapeutics as described in Box 1 can pave the way for innovative treatments in age-related neurodegenerative disorders. Currently, more than 20 senolytic trials are under way, with four focusing on CNS conditions. Simply stated, the true potential benefits of these interventions are yet to be determined. While acknowledging the challenges involved with advancing this new class of treatments to the CNS, we maintain optimism and enthusiasm for the future of senotherapy development and its application to nervous system health. Addressing these challenges not only holds promise for improving the quality of life for ageing individuals but also has the potential to alleviate the broader societal and economic impacts of an ageing population.

Glossary

Cell cycle arrest

The interruption or halting of the cell cycle at specific checkpoints to prevent the completion of cell division

Immunosenescence

The ageing-related decline in efficacy of the adaptive and innate immune systems, leading to increased susceptibility to infections, reduced vaccine efficacy, less efficient clearing of senescent cells and a higher incidence of age-related diseases

Oxylipin

A large class of bioactive lipid metabolites derived from the oxidation of polyunsaturated fatty acids

Phagoptosis

A form of cell death characterized by the removal of living cells by phagocytic cells, such as macrophages and microglia

Progeroid mice

Genetically engineered or naturally occurring mouse models that exhibit accelerated ageing processes similar to those seen in human progeroid syndromes

PROTAC senolytics

Proteolysis-targeting chimeras that enable the targeted degradation of specific pro-survival proteins within senescent cells

SASP index

A quantitative composite score of cellular senescence that measures the expression of senescence-associated secretory phenotype (SASP) components using techniques such as quantitative PCR, enzyme-linked immunosorbent assay or mass spectrometry

Senescence-associated secretory phenotype (SASP)

The complex molecular milieu secreted by senescent cells affects neighbouring cells through extracellular matrix remodelling, chronic inflammation, attracting immune cells and altering function. SASP factors include cytokines, chemokines, growth factors, proteases, bioactive lipids, exosomes and free non-coding nucleic acids

Senescence-messaging secretome (SMS)

Secreted proteins that contribute to the induction and maintenance of senescence; later updated to reflect a greater variety of molecules and renamed the senescence-associated secretory phenotype