Neuronal senescence may drive brain aging

Senescence is a programmed cellular response that can be activated after stress and culminates in a distinctive change of cell state. When faced with irreparable damage, especially during disease or chronic aging-associated insults, many cells either undergo programmed cell death (apoptosis) or shift to an apoptosis-resistant nonproliferative senescent state. Understanding the conditions under which cells senesce has changed from a narrow response to proliferative exhaustion to a diverse, general strategy to manage molecular damage; mobilize countermeasures; and preserve tissue structure, function, and organismal health. However, in older individuals senescent cells accumulate to the extent that they contribute to tissue dysfunction. In particular, recent studies have shown that neurons undergo senescence despite being postmitotic and incapable of experiencing the hallmark proliferative arrest. Neuronal senescence has been largely observed in cases of disease and neurodegeneration, potentially placing senescent neurons at the threshold between healthy aging and disease in the human brain.

Neuronal senescence is a state shift induced primarily by various forms of aging-associated stress—including DNA damage, reactive oxygen species (ROS), or oncogenic signaling—that results in a neuron that is metabolically active but information processing deficient. The initial stage of stress response in neuronal senescence eventually proceeds to a mature senescent state, with epigenetic and morphological restructuring, metabolic reprogramming, and the development of an inflammatory mixture of cytokines and chemokines known as the senescence-associated secretory phenotype (SASP) that can activate nearby cells, including astrocytes and microglia. The discovery of senescence in neurons, a cell type that is postmitotic since birth, has been surprising given that proliferation arrest was long considered central to the senescent state.

Neuronal populations in the human brain are diverse and complex, and there are ongoing efforts to develop a complete brain cell atlas. Given this diversity, are there particular subtypes of neurons that are vulnerable to senescence? The most often reported subtypes are excitatory neurons from cortical areas, specifically the prefrontal cortex. Senescent neurons are typically observed in neurodegenerative diseases, primarily Alzheimer’s disease (1–4) but also in other conditions, including progressive supranuclear palsy (4) and hyperinsulinemia (5). In an in vitro model that used aged human cortical neurons, excitatory glutamate-expressing cells senesced more frequently than inhibitory γ-aminobutyric acid (GABA)–expressing cells (3), and single-cell profiling of the human brain identified cortical excitatory neurons as the most prevalent neuronal subtype to senesce (2). Cortical neurons also senesce readily during normal aging, without a disease component, as has been observed in rat cortical neurons (6), aged mouse pyramidal neurons (7), and aged mouse calbindin-expressing neurons (8).

Diseases that affect more specific subpopulations of neurons can manifest a senescent phenotype, as in human induced pluripotent stem cell (iPSC)–derived dopaminergic midbrain neurons in Parkinson’s disease (9, 10) or neurons surrounding an acute physical injury site in humans and mice (11, 12). Senescence in the cerebellum is less clear, with senescence being observed during aging or DNA damage in mice (8, 12), but these cells are largely absent from brain-wide efforts to identify senescent neurons in humans (13). The colocalization of senescent neurons with disease-affected regions suggests that pathogenic mechanisms initiate from and/or promote neuronal senescence. Given the relatively low number of studies investigating neuronal senescence, it is likely that other aging-associated stressors could trigger senescence in nonexcitatory or cortical subtypes.

A major challenge is the establishment of a specific, universal marker of senescence. Owing to the inherent complexity of the senescence phenotype and the myriad cell types that can senesce, the current accepted strategy is a multimarker approach. This approach has resulted in a wide variety of phenotypes being classified as markers of senescence in neurons, some of which are observed in states other than senescence.

One of the most common markers of neuronal senescence is expression of cyclin-dependent kinase inhibitors, including p21, p16, and p19. p21 is expressed in senescent iPSC-derived neurons in vitro from individuals with Parkinson’s disease, spinal cord injury, and hyperinsulinemia, as well as in mouse neurons that contain DNA double-strand breaks (DSBs) in vivo and after prolonged in vitro culture of rodent and human neurons (1, 5, 6, 9–11). Initially high p21 expression in human senescent neurons becomes lower in late senescence when other features such as morphological changes and a SASP develop, highlighting the temporal heterogeneity of the phenotype (3). Expression of p16 is frequently (3–5, 7, 12, 13) but not necessarily (1, 9, 10) observed in neuronal senescence in mice and human iPSC-derived neurons. p19 has been reported in human neuronal senescence in Alzheimer’s disease (13).

Markers for DNA damage are also prominent in neuronal senescence, as evidenced by the presence of the DSB marker phosphorylated histone H2AX (γ-H2AX) (1, 4, 6, 8), changes in nuclear size (3, 9, 13) or structure (6, 9, 13), and epigenetic restructuring and formation of senescence-associated heterochromatin foci (6, 8). Lysosomal acidification and lysosomal senescence-associated β-galactosidase have been reported in many neuronal senescence studies (3, 5, 6, 8–11), although the ultimate reliability of these as markers of neuronal senescence is unclear owing to inherent differences in lysosomal biology between neurons and proliferative cells.

Neurons are extremely long-lived cells that have a high metabolic demand, which is accompanied by ROS production. These metabolic requirements necessitate a reliance on autophagy to recycle oxidized macromolecular structures and organelles. Thus, impaired autophagy can lead to accumulation of ROS-induced damage and a subsequent senescence response in rat neurons in vitro (6). Conversely, lack of autophagy-induced suppression of GATA-binding protein 4 (GATA4) expression, a transcription factor involved in innate immune signaling, can promote a neuronal SASP (7). The metabolic demands of neurons also make them vulnerable to senescence through mitochondrial mechanisms (3–5, 9, 10). Recent studies of an aged human cell culture model identified impaired oxidative phosphorylation tied to senescence in aging human neurons (3, 14), and impaired glucose metabolism culminates in reactivation of cell cycle machinery followed by neuronal senescence in mice (5). Consistently, the uncoupling of oxidative phosphorylation from electron transport and increased ROS have been proposed as markers of neuronal senescence (8), and mitochondrial dysfunction is a core component of Parkinsonian neuronal senescence (9, 10).

Because neuronal senescence is more common in the context of neurodegenerative disease than in healthy aging, and metabolic alterations could be an early phenotypic change in neurodegenerative pathogenesis, it seems plausible that senescence constitutes a compensatory mechanism to endure a suboptimal or high ROS–producing metabolic environment. Under this vulnerable state, increasing biosynthetic glycolytic intermediates or oxidative damage could induce neurons, a cell type with particularly high energetic demands, to senesce. Long-term senescence could then result in neurodegeneration from SASP release and the accumulating effects of irreversible neuroinflammation on surrounding cells (see the figure).

Aging-associated stress triggers senescence.

Neurons undergo terminal differentiation from a neural stem cell early in life followed by permanent cell cycle arrest. As a result, neurons experience a lifetime of age-associated stress, which culminates in a subpopulation with a senescent phenotype late in life. Despite being nonproliferative, senescent neurons present other common phenotypes associated with senescence.

What evidence is there of a neuronal SASP that affects neighboring cells in brain disease pathogenesis? In aged postmortem human brains, p16-expressing neurons and glia cluster most significantly around p16-positive mature neurons (3). Many SASP factors attract immune cells, and microglia have been reported to be activated by senescent neurons in human iPSC–derived Parkinson’s disease neurons in vitro or DSB-containing neurons in mice (1, 10). Microglia also had reduced activation after selective chemical clearance of senescent cells (senolysis) in an Alzheimer’s disease mouse model (4). Further, media taken from cultures of senescent neurons triggered a reactive phenotype in astrocytes (3) and induced premature senescence in primary rat neurons in vitro (6). Increased numbers of senescent glia have also been observed during brain aging and disease, and neuronal SASP could be a source of senescence-triggering stress in these cells.

The SASP factors reported in neurons are diverse and likely differ according to the temporal stage of senescence, cell type, model system, and the nature of the initiator of senescence. Interleukin-6 is a widely observed neuronal SASP factor, including in human Alzheimer’s disease cerebrospinal fluid and neurons in vitro, aging rat neurons ex vivo, neurons after stroke in humans and mice in vivo, and from mouse neurons with DSBs (3, 6–8, 12). However, it is absent from Parkinsonian human iPSC–derived neurons, hyperglycemic mice, and mouse spinal injury models of neuronal senescence (5, 9, 11). Various chemokines are also involved in neuronal SASP. C-C motif chemokine ligand 2 (CCL2) is the primary SASP factor in DSB- and autophagy-induced neuronal senescence (1, 6), and its expression is reduced after senolysis (11). One of the key accumulating transcription factors in autophagy-induced neuronal senescence, GATA4, is implicated in a separate study of DNA damage–induced neuronal senescence (7), suggesting that CCL2 (a GATA4 target) could be released in vivo in mice after DNA damage as well. A cell-intrinsic immune response appears to be a likely mechanism by which a neuronal SASP is maintained, either through DNA damage response signaling (1), GATA4 activity (7), or nuclear factor κB (NF-κB) signaling (4). However, many other SASP factors likely remain to be discovered, and it will be important to identify the key features that mediate immune or glial cell activation, which likely plays a role in pathogenesis.

Senescent cells are a major target for ameliorating age-associated decline. Could senescent neurons be targeted by therapeutics to treat brain diseases? Senescent neurons are sensitive to senolytics, which are a diverse class of drugs that selectively induce apoptosis in senescent cells. For example, the combination of dasatinib and quercetin killed senescent cells, including human Alzheimer’s disease and Parkinson’s disease neurons in vitro and neurofibrillary tau-containing neurons in aged transgenic mice (3, 4, 9). A recent clinical trial of this combination, however, showed low brain penetrance of quercetin, which limits efficacy (15). The penetrance of many senolytics is yet to be tested, but those larger than 500 Da are unlikely to cross the blood-brain barrier (BBB). Other drugs have shown promise as senolytics of senescent neurons. In a hyperglycemic mouse model, liraglutide, a type 2 diabetes drug that modulates insulin release, reduced the numbers of senescent neurons (5); the B cell lymphoma 2 (BCL-2) inhibitor ABT-263, which promotes apoptosis, cleared senescent neurons after spinal injury in mice (11); and the autophagy-promoting azithromycin or antioxidant flavonoid fisetin eliminated senescent dopaminergic human iPSC–derived neurons in vitro (9). Aged Alzheimer’s disease patient–derived neurons were more sensitive to clearance by the BCL-2 inhibitor ABT-737 in vitro than were neurons from cognitively normal individuals (14), but healthy iPSC-derived dopaminergic neurons are also affected by ABT-737 in vitro (9). This underscores the need for future work to fully define the pathways that senescent neurons rely on for survival so that they may be specifically targeted.

Neuronal senescence is also reduced in mice indirectly, including through dietary restriction (8) or administration of the autophagy-promoting disaccharide trehalose (6), both of which could be considered senomorphic (agents that can reverse a senescent phenotype) and an alternative strategy to the ablation of senescent neurons, which are irreplaceable. An important consideration for in vivo studies of senolytics and senomorphics is the contribution of direct or indirect effects of these drugs on outcomes because clearance of non-neuronal senescent cells could explain some benefits. Defined in vitro models could be a useful method for testing the direct effects of senotherapeutics (3). Progress has been made in this area, including transdifferentiation strategies to maintain age, introduction of genetic variants that promote senescence, and accelerated aging through drug treatments or progerin mutations.

Evidence collected on neuronal senescence thus far points to a causative role for persistent senescent neurons in triggering pathogenesis in the aged brain. Therefore, accurately identifying and targeting senescent neurons represents an opportunity for intervening in age-associated diseases and an area that would benefit from continued investigation of genetic markers of neuronal senescence. Further, the exact features of neuronal senescence that are pathogenic are still not fully resolved, and future work should address which changes are the most consequential or if any provide physiological benefits. In addition to in vivo developments, new in vitro techniques that produce aged human postmitotic cells constitute exciting new models for studying and testing interventions that target senescence. Future investigation and understanding of neuronal senescence will feed into the development of more advanced and precise senotherapeutics, a timely event considering the rapid aging of many societies and the immense need for therapeutic interventions for age-associated diseases. ■