More than a number: Incorporating the aged phenotype to improve in vitro and in vivo modeling of neurodegenerative disease

,


Global ageing: Implications for neurodegenerative disease
Substantial scientific advancements have resulted in longer life expectancies.Global average life expectancy in 2019 was 9.8 % higher than in 2000 and further increases are predicted, resulting in a significant aged population (Cao and Ho, 2020).For example, in Sweden, more than 50 % of the 50,000 babies born in 2007 are expected to become centenarians, compared to ~750 people who turned 100 in 2007 (Vaupel, 2010).Concerningly, however, healthy average life expectancy (HALE) has not maintained this pace.In 2019, global HALE was 63.7, compared to the overall life expectancy of 73.4 years (Cao and Ho, 2020).This means that, while individuals may be living longer, they are not necessarily maintaining their health for this full period, resulting in increased prevalence of age-related diseases and demand for healthcare services (Chang et al., 2019).Particularly, incidence of neurodegenerative diseases (NDs), including Alzheimer's disease (AD) and Parkinson's disease (PD), which are highly prevalent in the >65 population, has increased dramatically (Hou et al., 2019).This is significant, as these age-related NDs have long-term cumulative societal and economic impacts (Arafah et al., 2023).Analysis of the global burden of disease study 2019 revealed AD was responsible for 25.3 million disability adjusted life years (DALYs) and PD 6.2 million in 2019 (Ding et al., 2022).This represents a 161 % and 128 % increase in DALYs caused by AD and PD, respectively, since 1990 (Ding et al., 2022).Additionally, AD has an annual treatment cost of $305 billion USD (Wong, 2020) and PD around $52 billion (Yang et al., 2020a), with costs set to rise with the number of diagnoses.Current management focuses solely on symptomatic relief, as disease-modifying therapeutics are lacking (Kalia and Lang, 2015;Yiannopoulou and Papageorgiou, 2020).While symptomatic relief is critical for improving patient's daily lives, this does not address the underlying brain mechanisms of the disease, ultimately resulting in increased healthcare burden as the disease continues to progress and comorbidities arise (Arafah et al., 2023).
To develop disease-modifying treatments for NDs, there needs to be a better understanding not only of disease-specific pathophysiology, but also, the shared contribution of ageing across NDs (Mayne et al., 2020).Age is the primary risk for developing NDs (Hou et al., 2019), and it is well documented that age-related cellular changes can impact physiological function and exacerbate disease pathology (González-Gualda et al., 2021).Despite this, while there are exceptions (DiMarco et al., 2023;Muñoz-Manchado et al., 2016), many preclinical ND studies are still conducted in young animals and in vitro studies are rarely carried out on the background of ageing.While this may be appropriate for certain study designs, for example where the focus is specifically on finding potential prophylactic therapeutic options for NDs (Yang et al., 2020b), this is nevertheless a considerable shortcoming for many investigations in the field.Assuming responses in young cohorts are reflective of older populations is inaccurate (Sun et al., 2020) and has the potential to impact translation of findings from preclinical therapeutic investigations to aged patients.With this in mind, here we first introduce the aged phenotype in mammals and then discuss how aspects of this phenotype may be better incorporated into both in vitro and in vivo modelling of NDs.

The aged phenotype
As ageing characterised by decline in function and is often described as a pathophysiological process, there is debate as to whether healthy ageing truly exists (Faragher, 2015).For this review, healthy ageing refers to the biological changes that occur in an organism later in life, after the growth and development period has ended and advanced adulthood has begun, and which are not associated with a diagnosed condition.There are a number of visible features commonly associated with natural ageing (Fig. 1A) which define an individual as aged.These include hair greying, skin changes (Halliwell and Dittmar, 2003) and musculoskeletal changes, including muscle wasting, joint immobility and decreased bone density (Azzolino et al., 2021).Functionally, ageing corresponds to changes in multiple systems, including digestive (Bhutto and Morley, 2008), reproductive (Kaufman et al., 2019;Shirasuna and Iwata, 2017), circulatory (Strait and Lakatta, 2012) and immunity (Wang et al., 2022), with the latter two contributing to increased vulnerability to infections and diseases (Strait and Lakatta, 2012).Cognitive function is also affected with natural ageing, with declines Fig. 1.Biological ageing A) whole-body distinguishing features including changes in (1) hair and skin (2) muscle, (3) cognition, (4) vision, (5) digestion, (6) reproduction, (7) bones and joints, (8) circulation and (9) immunity.B) cells display aged/senescent features including (1) increased DNA damage and decreased DNA repair, (2) telomere shortening, (3) increased lysosomal presence and activity, (4) decreased proteasome activity causing an increase in intracellular protein aggregates, (5) increased production of reactive oxygen species (ROS), (6) mitochondrial damage, (7) increased mitochondrial DNA mutations and (8) the development of the senescence-associated secretory phenotype (SASP).C) functional changes to inflammation and metabolism driven by cell signalling changes with an increase in the DNA driven immune response influenced by the (1) cGAS-STING pathway, (2) AIM2 pathway and (3) the TLR9 pathway which all contribute to (4) increased inflammation termed 'inflammageing'.Metabolic signalling pathways are also altered with (5) altered insulin/IGF-1 binding and (6) mitochondrial damage both influencing changes in the (7) mTOR and (8) AMPK pathways.Both pathways can regulate transcription of factors important for cell cycle regulation, metabolism, and apoptosis such as (9) FoxO.
seen in several domains, including memory, processing speed and reasoning ability, although these changes are not as pronounced as mild cognitive impairment or AD (Moreira et al., 2019).Conversely, other aspects of cognition, such as vocabulary, are resistant to the effects of ageing and even increase until individuals are 60 + years old (Hartshorne and Germine, 2015).Given the multiple systems impacted, ageing is clearly multifactorial, reflected by several theories being put forward to explain mechanisms of ageing (reviewed: (da Costa et al., 2016)).This has led to the formulation of 12 hallmarks of ageing by López-Otín and colleagues (López-Otín et al., 2023): genomic instability, telomere shortening, epigenetic alterations, loss of proteostasis, disabled macroautophagy, dysregulated nutrient sensing, mitochondrial dysfunction, cellular senescence, stem cell exhaustion, alterations in intracellular communication, chronic inflammation and dysbiosis.While an in-depth consideration of all of these hallmarks is outside the scope of the current work, below we discuss markers of ageing in the context of a senescent phenotype specifically those relevant to brain ageing.We direct the reader to López-Otín and colleagues excellent review on the topic (López-Otín et al., 2023) for further reading on the additional markers not discussed here.The markers discussed in aged cellular phenotype below are critical to take into account for age-related neurodegenerative diseases.
Central to ageing mechanisms are the defining features of a senescent cell, (Fig. 1B) with cellular senescence named as one of the hallmarks of ageing (López-Otín et al., 2023).Senescent cells are often physically distinguishable, with enlarged and flattened morphology compared to young cells (González-Gualda et al., 2021).A characteristic feature of cellular senescence is cell cycle arrest, with loss of proliferative capacity.Cellular senescence is commonly assessed by evaluating levels of cell cycle regulators, including P16 and P21, kinases which play crucial roles in the transition from G1 to S phase of the cell cycle and are upregulated in G1-phase cell cycle arrest (Serrano et al., 1997).DNA damage, measured by DNA double strand breaks (DSBs), and telomere shortening (Fig. 1B (1-2)) are considered major drivers of cell cycle arrest and consequent senescent phenotype (Fumagalli et al., 2012), and accordingly were listed as two of the 12 hallmarks of ageing (López-Otín et al., 2023).Normally, the DNA damage response (DDR) counteracts DNA damage by either repairing damage or initiating apoptosis if the damage is beyond repair (reviewed: (Chatterjee and Walker, 2017).However, with ageing, rate of DNA damage overwhelms the capacity of the DDR, resulting in accumulated damage in both mitochondrial (mtDNA) and nuclear DNA, with subsequent altered gene expression, impaired cellular function and senescence (Victorelli et al., 2023).
Inflamm-aging was a term first devised by Franceschi et al in 2000 to describe the concurrent decline in immune function and development of a chronic pro-inflammatory state with increasing age, in the absence of any external immune stimulation (Franceschi et al., 2000) and has since become a prominent theory of ageing (Xia et al., 2016).Fittingly, chronic inflammation is described as a hallmark of ageing (López-Otín et al., 2023) and is thought to be driven by the development of the senescence associated secretory phenotype (SASP) (Olivieri et al., 2018) (Fig. 1B (8)), which is a key element of senescence discussed by López-Otín and colleagues (López-Otín et al., 2023).The SASP was observed to occur in multiple cell lines and in vivo in 2008 as a response to DNA damage (Coppé et al., 2008).This phenotype features high ongoing production of pro-inflammatory cytokines, such as interleukins (IL-6, IL-8, IL-1β) and Tumour Necrosis Factor (TNF, formerly known as TNF-α (Grimstad, 2016)), which contribute to further immune activation (Coppé et al., 2008).In the brain, microglia are thought to be responsible for the emergence of this damaging inflammation in NDs and the SASP (Hu et al., 2021), with specific microglial pathways implicated in this process.
The shift toward an inflammatory phenotype in the brain is associated with changes in major signalling pathways involved in DNA-driven immune recognition responses (Gulen et al., 2023) (Fig. 1C).For example, cyclic GMP-AMP synthase (cGAS) is a cytosolic DNA sensor, which, upon binding to DNA in the cytosol, initiates cyclic GMP-AMP (cGAMP) to bind with the stimulator of interferon genes (STING) located on the endoplasmic reticulum (Fig. 1C (1)).This binding induces a signalling cascade leading to production of type-1 interferons and proinflammatory cytokines (Sun et al., 2013).During ageing, cGAS recognition of cytosolic DNA danger signals increases, potentially due to increased nuclear and mitochondrial damage, as outlined in Fig. 1B (1-2 and 5)) (Gulen et al., 2023).The resulting increase in activation of this pathway is hypothesised to drive the ongoing low-grade inflammatory state seen in ageing (Gulen et al., 2023), a state which is also linked to neurodegenerative diseases (Guzman-Martinez et al., 2019) (Fig. 1C (4)).
Similarly, dysregulated intracellular metabolic processes are one of the hallmarks of ageing (López-Otín et al., 2023) and thus are critical in cellular and systems ageing.Key regulators of metabolism are insulin and insulin-like growth factor-1 (IGF-1), which both decline with age, as does binding to insulin/IGF-1 receptors (Fig. 1C (5)) (Toth et al., 2022).Not only does decreased insulin/IGF-1 signalling impair cerebral blood flow and contribute to neuronal dysfunction (Toth et al., 2022), it also causes dysregulation in the mechanistic target of rapamycin (mTOR) and AMP-activated protein kinase (AMPK) metabolic pathways (Yang et al., 2018)(Fig. 1C (6-8)).The mTOR and AMPK pathways are major regulators of cell metabolism, mitochondrial function, cell growth and cell survival (Kazyken et al., 2019), and of the many signalling pathways involved in metabolism, these are the best described and most closely linked to biological ageing (Morita et al., 2013).With age, there is a loss of sensitivity of the AMPK pathway, which normally has an inhibitory effect on mTOR signalling (Gwinn et al., 2008).This subsequently leads to an increase in mTOR signalling, which has been linked to the manifestation of age-related disorders (Bitto et al., 2016).Additionally, both AMPK and mTOR pathways induce transcription of genes related to mitochondrial function, in particular transcription of forkhead box class 0 (Fox0), a regulator of mitochondrial homeostasis (Davila et al., 2012).Dysregulation of these pathways with age may thus further exacerbate mitochondrial dysfunction.Interestingly, there is conflicting evidence regarding whether activation of the mTOR and AMPK pathways confers neuroprotective or damaging effects in ND (reviewed: (Querfurth and Lee, 2021)).Overall, however, it is evident that ND pathophysiology does not occur in isolation, but rather is concomitant with the complex and multifactorial physiological changes of ageing.It is thus imperative to consider these changes when modelling ND preclinically.

The aged phenotype may not only affect pathophysiology, but also directly influence drug response
Given that there is no single causative gene or pathway controlling ageing, but rather many changes contributing to the ageing process, results obtained in "young" models may not be translatable within an aged context.In fact, age-related differences in drug responses are widely recognised, with adverse drug reactions in elderly patients being twice as common compared to younger patients, particularly with drugs targeting the CNS (Beijer and de Blaey, 2002).
Further, age-associated alterations have been reported for multiple (but not all) members of the cytochromes P450 family of drugmetabolising enzymes (CYPs) (reviewed: (Kinirons and O'Mahony, 2004)), suggesting that drug pharmacokinetics are altered in natural ageing.In line with this, a study that looked at the mRNA expression profile of 101 xenobiotic-processing genes in mice found that 44 % of genes were altered with age in male mice and 64 % changed with age in female mice (Zidong Donna et al., 2012).Interestingly, although it would be reasonable to hypothesise that these changes would be further exacerbated in NDs, to date, this has yet to be explored in the literature (Wu and Lin, 2019).
Although CYPs are abundantly expressed and studied in the liver, the primary site of drug metabolism, CYPs are also widely expressed in the brain, with some forms, such as CYP2D4 and CYP2D18, expressed exclusively, or at much higher levels, here than in other tissues (Kawashima and Strobel, 1995;Komori, 1993).Importantly, brain CYPs play a role in metabolising many drugs that act on the CNS (for review, see (Miksys and Tyndale, 2013)) and can alter response to CNS-acting drugs (Khokhar and Tyndale, 2011).This may have implications for the treatment of CNS disorders, including NDs.Intriguingly, there is great inter-individual variation in brain expression levels of CYP levels, and this can be affected by multiple factors, including age.For example, in humans, brain levels of the neuroprotective enzyme CYP2D6 gradually increase with age, reaching their highest levels in those aged 65+ (Mann et al., 2012).This gives further support to the idea that the response to centrally acting drugs may differ in younger versus older adults.This response may be further altered in those with NDs.In line with this, those with PD were shown to have ~40 % lower levels of CYP2D6 within several brain regions, including the frontal cortex, cerebellum and hippocampus (although, interestingly, not in the substantia nigra or caudate), compared to age-matched controls (Mann et al., 2012).
These effects may be due, at least in part, to the increases in inflammation seen both with natural ageing (i.e.inflammageing) (Franceschi and Campisi, 2014) and in NDs (Guzman-Martinez et al., 2019).In support of this, pro-inflammatory factors are known to correspond to reduced activity of many drug metabolising enzymes (Aitken et al., 2008), including CYPs (reviewed: (de Jong et al., 2020)).A systematic review of 218 studies in adults by Lenoir et al. (2021) demonstrated that inflammation modulates the activity of multiple CYPs and thus alters drug pharmacokinetics, although the effect is dependent on the particular CYP isoform and source of inflammation considered (Lenoir et al., 2021).While the exact implications of this for understanding drug response in younger versus older individuals, as well as in individuals with a ND, remain to be explored, it nevertheless highlights that when testing potential therapeutics for NDs, either in in vitro or in vivo models, it is critical to account for the potential impact of age, in order to improve the translational relevance of the findings.

Laboratory models of the aged phenotype
The mechanisms underpinning ageing are multifactorial and thus make modelling age-related NDs complex.In the following section, key laboratory models of ageing are described and critically examined for their applicability to NDs.

Cells derived from aged donors
In vitro studies are a critical tool in understanding how the aged brain functions at the cellular level.Primary cells from aged individuals are understandably regarded as one of the most accurate models, as they can capture the features of an aged cell and retain the genetic characteristics of the donor when harvested using specific techniques.However, for the human CNS, obtaining such cells directly can be difficult, relying on cultures produced from post-mortem brains or tissue removed during brain surgery.The advent of human induced pluripotent stem cells (iPSCs) in 2006 (Takahashi and Yamanaka, 2006) resolved some of these difficulties, allowing somatic cells to be induced to become pluripotent by expression of four reprogramming factors known as OKSM factors or Yamanaka factors (Takahashi and Yamanaka, 2006).These iPSCs can then be differentiated into CNS cells and used to model neurological disorders.
Using this method, both neurons (Miller et al., 2013) and glial cells (Abud et al., 2017;Voulgaris et al., 2022) derived from iPSCs have successfully been used as models of AD and PD (reviewed: (Pandey et al., 2022)).For example, utilising iPSCs derived from somatic cells of patients with familial PD, it is possible to generate dopaminergic neurons that express pathological PD mutations and show PD-like pathology in culture, including increased pathological α-synuclein aggregates (Kouroupi et al., 2017), neurite degeneration (Sánchez-Danés et al., 2012) and dysregulated dopamine pathways (Jiang et al., 2012).Cultures derived from iPSCs obtained from individuals with idiopathic PD also display PD-like impairment of neurite formation (Sánchez-Danés et al., 2012) and increased autophagy and mitophagy (Hsieh et al., 2016).Similarly, neurons derived from iPSCs of AD patients carrying the APP mutation display short fragments of amyloid-beta (Aβ) and increased Aβ42:40 ratios.Importantly, however, when using iPSCs to model ageing, studies have shown that the developmental reset induced by treatment of the somatic cells with Yamanaka transcription factors erases senescent features in these cells (Kim et al., 2018;Tang et al., 2017).Table 1 summarises several studies where reset of the ageing phenotype has been demonstrated in iPSC-derived neurons and glia.Reset of telomeres, heterochromatin levels, mitochondrial health, nuclear organisation, and protein damage is widely thought to occur during induction to pluripotency (Miller et al., 2013) and has been reported across multiple studies on ageing iPSCs (Lapasset et al., 2011;Ocampo et al., 2016) (Prigione et al., 2011;Suhr et al., 2009).While iPSC technology has great potential for regenerative therapy in ND treatment (Limone et al., 2022), simply utilising iPSCs derived from older individuals may be insufficient to accurately replicate the aged phenotype of CNS cells.
Alternative to iPSCs, directly induced neurons (iNs) aim to achieve control over the phenotype of the induced cells.These are derived from somatic cells by lentiviral or microRNA-based expression of "BAM" factors and neurogenesis factors, such as NeuroD1 and Ngn2, similar to the expression of Yamanaka factors in the reprogramming iPSCs, but bypassing the need to become pluripotent before differentiation (Vierbuchen et al., 2010).Promisingly, these cells display retention of age after differentiation, with epigenetic age of the original donor cell, measured by DNA methylation, displaying a high correlation with epigenetic age of the iN (Huh et al., 2016).Additionally, iNs derived from idiopathic PD patients maintain age-associated genetic signatures and α-syn pathology, which is not present in iPSC-neurons derived from these same patients (Drouin-Ouellet et al., 2022).Similarly, directly converted neural progenitor cells display ageing when further differentiated to astrocytes, with changes summarised in Table 1 (Gatto et al., 2021).Thus, skipping induction to pluripotency through direct conversion appears to avoid the senescence reset, allowing for a more robust model of ageing (Huh et al., 2016), as well as retention of sporadic neurodegenerative pathology (Drouin-Ouellet et al., 2022).While promising, protocols for direct cellular reprogramming into brain cells ↑SA-B-gal (Boraldi et al., 2015;Caldeira et al., 2014;Deng et al., 2023;Kurz et al., 2000), ↑lysosomal mass (Kurz et al., 2000) Activated morphology, ↑Matrix Metalloproteinase (MMP)-2, ↓NF-kB, ↓MMP-9, ↓TLR-2, ↓TLR-4, ↓autophagy, ↓migration and phagocytic capacity (Caldeira et al., 2014) Enlarged cells, vacuolated cytoplasm, aHDF: longer growth periods than nHDF ( Boraldi et al., 2015) Enlarged nuclei, ↓proliferation, ↑P16/P21, ↑yH2AX, ↑uPAR, SASP (Deng et al., 2023) Ageing have been less extensively described than those for iPSCs.Additionally, there is a need for further validation regarding the neuronal phenotypes induced from differing cellular origin (Chouchane et al., 2017) and potential risks involved in using iNs for regenerative therapy (Kim et al., 2021).Overall, it is important to note that the donor age does not consistently affect the lifespan of primary cells in vitro (Cristofalo et al., 1998;Kaji et al., 2009).This is exemplified from cultures generated from participants in the Baltimore Longitudinal Study of Ageing.This study compared the proliferative capacity of fibroblasts collected from young donors (20-29 years) versus old donors (65 + years) both crosssectionally and longitudinally, with donor matched biopsies up to nine years apart (Smith et al., 2002).No difference in lifespan of fibroblasts was observed cross-sectionally, except when derived from female donors (Smith et al., 2002), where a reduced proliferative potential correlated with donor age.Comparison of cells from the same donor longitudinally demonstrated a trend to decrease in proliferative potential over time, but this was not statistically significant, potentially on account of considerable variation between cell lifespan from different donors (Smith et al., 2002).Conversely, a study comparing fibroblasts from donors aged 21-36 versus donors aged 63 + reported significant effect of donor age on the proliferative potential of cells in culture (Schneider and Mitsui, 1976), although large variability and overlapping data points between age groups was observed.These studies highlight that in vivo chronological age may not directly translate to reduced proliferative capacity and senescence in vitro, with variation between donors, combined with variations in cell isolation and culture protocols, leading to challenges in reproducibility (Qadan et al., 2018).While simply generating iPSCs or iNs from aged donor cells may not fully encapsulate an aged phenotype in the sense of proliferative capacity, the potential utility of these cells in ND research and therapy should not be understated.Further, other markers in addition to proliferation can be used to assess age and determine the age of a culture, as discussed in more detail below.

Replicative senescence
Other than utilising cells from aged donors, the most prominent method of inducing senescence in vitro is by culturing cells to a proliferative limit, termed replicative senescence, first proposed by Leonard Hayflick in 1965(Hayflick, 1965).In these early studies, primary cells were cultured from human foetal tissue continually for up to 11mths or 55 passages.The "Hayflick limit" is widely accepted and is the basis for characterising serially cultured cells as a model of cellular ageing (Phipps et al., 2007).Passaging cells to this limit produces the characteristic cell cycle arrest and SASP seen in senescent cells, without researchers intentionally altering specific cellular pathways (Hayflick, 1965) and thus may result in a more translationally relevant aged phenotype.In line with this, studies using peripheral cells (presented in table 1), including dermal fibroblasts and umbilical vein endothelial cells, display a senescent phenotype featuring the SASP (Deng et al., 2023), SA-β-gal (Boraldi et al., 2015;Caldeira et al., 2014;Kurz et al., 2000), increased cell cycle regulators (Deng et al., 2023), decreased proliferation (Boraldi et al., 2015) and increased markers of DNA damage (Deng et al., 2023), all key components of the aged phenotype (Fig. 1).When looking at CNS cells, similar results have been observed; however, the use of CNS cell types in studies of replicative senescence is sparce.Primary microglia have been shown to change from a reactive to a chronically active, age-like phenotype (Table 1) when kept in culture for a 16-day period (Caldeira et al., 2014).The lack of investigation of replicative senescence in neurons is due to the fact that they are postmitotic, and thus do not replicate and cannot undergo replicative senescence (Hartmann et al., 2023), making this method less applicable for the modelling of NDs.
Replicative senescence as an in vitro model of ageing has been validated mostly in passaged primary cells, where replicative senescence occurs more readily than in immortalised cell lines.Recently, however, untransformed primary mouse T-cells have demonstrated the ability to continue to divide and remain functional in culture for 10 years, potentially challenging the idea of universal applicability of the Hayflick limit (Soerens et al., 2023).Additionally, a primary cell model of ageing is difficult to consistently reproduce due to variation between cell populations, presenting a challenge for widespread use and high throughput industry applications.Furthermore, there is limited capacity to conduct complex experiments or investigate long-term effects, due to the limited time primary cells can be sustained in culture (Hayflick, 1965).Alternatively, investigating passage number as a proxy for ageing in immortalised cell lines may offer the opportunity to study phenotypic or functional changes over time, rather than simply modelling cells at their proliferative end (Sapieha and Mallette, 2018).Although immortalised cells have been argued as unfit for a model of ageing due to their limitless proliferative capacity and resistance to replicative exhaustion (Bunc et al., 2019), it is a recommended laboratory practice to only culture cell lines for a limited passage range, beyond which the cells are assumed to have diverged from the original cell phenotype (Geraghty et al., 2014;Hughes et al., 2007).Exactly what changes are induced in these cells with passage is not well characterised (Hughes et al., 2007), and, potentially, these changes may mirror some of the changes associated with the aged phenotype.Of particular relevance for NDs, it would be useful to characterise the phenotypic changes that occur in both immortalised glial cell lines (e.g.BV2 cells) and in neuronal-like cell lines (e.g., both differentiated and undifferentiated SH-SY5Y neuroblastoma cells, an in vitro model of PD (Xicoy et al., 2017)).If such immortalised cell lines do show alterations consistent with the aged phenotype with increased passage numbers, then this has significant implications for how these cells might be used to more appropriately model ND on an aged background in vitro.

Emerging methods and future directions
While single monolayer cultures provide a basic way to induce the senescence-related molecular changes associated with ageing, these models lack the complexity of co-culture and organoid brain models.This is particularly relevant for NDs, which are multifaceted, heterogenous diseases that involve interactions between multiple brain cell types (Luchena et al., 2022).In recent years, these alternative culture methods have been developed that provide an avenue to improve modelling of NDs.The use of co-culture models, which involve simultaneous expression of neuronal and glial cells, allows for examination of cellular interactions at the level of the tetrapartite synapse.Co-culture is being used to improve modelling of NDs and provide critical insights into cellular interactions in these diseases (see Table 1).For example, in a 2D triple co-culture model of murine astrocytes, microglia and neurons, exposure to oligomeric amyloid beta resulted in neuroinflammation and synaptic loss, providing a model that is more reflective of pathological processes associated with AD (Luchena et al., 2022).Similarly, using a co-culture model of LPS-stimulated macrophages and SH-SY5Y cells (model for dopaminergic neurons), reactive macrophages were shown to induce oxidative stress and consequent increased alpha-synuclein nitration and cell death in the SH-SY5Y cells (Shavali et al., 2006).
While allowing for complex interactions to be captured, it remains to be seen how the aged phenotype might be best incorporated into such co-cultures.One promising study by Bigagli and colleagues (2016) maintained primary neuronal/glial co-cultures up to 27 days in vitro and demonstrated that these cultures displayed multiple signs of cellular senescence, including increased levels of SA-β-gal, γ-H2AX, oxidative stress and pro-inflammatory cytokines (Bigagli et al., 2016).Interestingly, however, analysis of 180 microRNAs showed a pattern of change more consistent with transgenic TgCRND8 mice (model of amyloid-β deposition) than with physiologically aged mice, suggesting that longterm maintenance of the co-culture model may more closely resemble the changes of pathological ageing than natural ageing.Nevertheless, this provides preliminary evidence that aspects of the aged phenotype can be incorporated in more complex co-culture models.
Similar to co-culture models, brain organoid models demonstrate potential to replicate the complexity of the aged brain.Organoid models are stem cell derived 3D tissue cultures comprised of multiple cell types aimed at recapitulating the complexity of whole systems, in a way that co-culture and single monolayer culture cannot (Shou et al., 2020).These models are generated by embedding embryoid bodies into extracellular matrix and inducing neural differentiation by supplementation with growth factors to support growth of specific brain structures (Shou et al., 2020).For NDs, these complex models have demonstrated utility in modelling ND pathology.Brain organoids have demonstrated pathological Aβ aggregation and Tau phosphorylation when derived from familial AD patients with cells overexpressing APP and PSEN1 (Choi et al., 2014;Raja et al., 2016).Similarly, organoids which have the LRRK2-G2019S mutations seen in familial PD display α-synuclein aggregation and altered neuronal morphology, as is seen in PD (Kim et al., 2019).Excitingly, organoids may also offer a way to incorporate an aged phenotype into models of ND.In support of this, a recent study by Ao and colleagues (2022) demonstrated that 45-day-old brain organoids show upregulation of key ageing markers, including cell cycle regulators and pro-inflammatory cytokines, in response to monocyte infiltration from aged patients compared to cultures infiltrated with young monocytes (Ao et al., 2022).Therefore, while embryonic stem cells, from which organoids are developed, naturally have a young cellular phenotype, with elongated telomeres not reflective of the senescent phenotype seen in the aged brain (Ao et al., 2022), long-term culture of organoids may produce a more translationally relevant model for studies of ND.
In addition to current model limitations, another major challenge in reproducing the aged phenotype in vitro is the inconsistency of markers used across studies, which can make it difficult to fully evaluate models.This was recently addressed by Hartmann and colleagues (2023), who validated a biomarker-based panel to assess age in vitro in fibroblast cultures (Hartmann et al., 2023).This panel included many ageing markers discussed above, including DNA damage, telomere attrition, SA-β-gal, SASP, cell cycle arrest and morphological changes, as well as additional markers, such as histone modification and changes in nuclear lamina protein: Lamin B. Using these markers, an 'AgeScore' was derived, which correlated well with chronological donor age in both healthy individuals and donors with progeroid syndromes (diseases associated with accelerated ageing) (Hartmann et al., 2023).
While not yet investigated, this AgeScore may also be translatable to CNS cells.However, neuronal populations are post-mitotic, meaning cell cycle arrest, the defining feature of replicative senescence, cannot define an aged neuron (Hartmann et al., 2023).Despite this, a recent study investigating senescence 1wk following repeated mild traumatic brain injury induced by closed skull controlled cortical impact in C57BL/6 mice identified 15 clusters of neurons in injured mice, all expressing at least one marker of senescence (Schwab et al., 2022).This study provides evidence that, while neurons are post-mitotic, they can still demonstrate traditional markers of a senescent phenotype following traumatic brain injury, hypothesised to accelerate brain ageing (Cole et al., 2015).Thus, a modified version of the "AgeScore" may have utility even in evaluating neuronal cultures and subsequent standardisation of modelling ageing in vitro between studies.
Overall, while there is a multitude of research on inducing and characterising age in cell culture models (reviewed: (González-Gualda et al., 2021)), there has been less implementation of these models in ND research.This may be due to the difficulties in replicating results from primary culture across laboratories and across studies (Hirsch and Schildknecht, 2019) or the complexity of modelling disease alongside brain ageing.While research using standard models has and will continue to make significant contributions to the ND research field, given the critical role of age in the risk of ND development, incorporating aspects of the aged phenotype in vitro has the potential to lead to more translationally relevant models of ND.This will enhance our understanding of the pathophysiological mechanisms that drive ND emergence and lend insight into the therapeutic effects of putative treatments.

In vivo models of ageing: Potential applications to ND research 2.2.1. Spontaneous and accelerated models of ageing
To date, research incorporating aged animals has shown differences in motor performance, cognition and pathology following induction of neurodegenerative models.Recently, in an intracerebroventricularly injected streptozotocin model of sporadic AD, aged rats at 23 months old demonstrated increased p-Tau/Tau levels compared to controls whereas this increase was not displayed in young AD rats compared to controls (Gáspár et al., 2022).Additionally, these aged rats showed increased susceptibility to toxicity following induction of the model compared to young rats.Similarly, age impacted the effect of 6-hydroxydopamine lesion on skilled motor function and TH positive neuronal loss in the substantia nigra (Barata-Antunes et al., 2020).The results from these studies, and others discussed at length by Sun and colleagues (Sun et al., 2020), highlight the importance of including age in pre-clinical models of NDs in order to fully recapitulate a disease phenotype relevant to those seen in aged patients in the clinic.
A natural model of ageing would be most accurate in recapitulating all elements of this phenotype in vivo.The naked mole rat has a lifespan upwards of 37 years and low incidence of age-related neurodegeneration (Delaney et al., 2021).This species has provided insight into the mechanisms of natural ageing and disease resistance; however, due to their long lifespan, the utilisation of naked mole rats in laboratory research is impractical (Delaney et al., 2021).Additionally, given their resistance to age-related neurodegenerative pathology, genetic or experimental manipulation is required to make them a relevant model for ND research.Instead, for an estimated 80 % of ageing research, shorter lived rodent models are utilised (Andreollo et al., 2012), such as mice and rats, which live around 3 and 3.5 years, respectively, and, therefore, make it logistically feasible to house and obtain naturally aged animals within the time-frame of a reasonable research study (Andreollo et al., 2012).
Importantly, translating ageing across species is difficult due to differences in rate of maturation.In line with this, from birth to 1-month, mice mature approximately 150 times faster than humans, but this drops to 45 times faster in mice aged 1-6 months and to 25 times faster in mice aged 6-months or older (Flurkey et al., 2007).Despite this, some guidance has been put forward by Jackson Labs (Jackson Laboratory, 2020).In brief, based on survival analyses in a large cohort of C57BL/6J mice, the most commonly used inbred strain, the following life stages have been proposed: (1) mature adult from 3-6 months of age, roughly equivalent to 20-30 years of age in humans; (2) middle age from 10-15 months of age, which corresponds to approximately 38-47 years of age in humans and (3) old age from 18-24 months+, which is comparable to humans from 56 to 69 years of age (Jackson Laboratory, 2020).Interestingly, given that the old age period in mice is approximately equivalent to the period when age-related neurodegenerative diseases typically begin to arise in humans (Hou et al., 2019), while mice at 18-24 months old may not fully encapsulate the full aged phenotype of humans 70+ years of age, they can still provide valuable insights into the impact of ageing on disease pathophysiology and vulnerability to age-related diseases.It is important to investigate, however, how these age equivalents might vary as a factor of mouse strain, or indeed in other species, such as rats.In line with this, allele variation has been shown to result in both phenotypic differences (Simon et al., 2013) and disparities in response to an intervention (Liao et al., 2010), even in closely related strains.
Overall, the use of aged rodents comes with caveats, such as the presence of comorbidities, leading to many animals meeting ethical criteria for euthanasia prior to becoming aged (Brayton et al., 2012).While this may reflect population morbidities in humans, a portion of the study cohort is lost prior to reaching the experimental starting point  (Takeda et al., 1981) ↓spine density in hippocampus, ↑axonal dystrophy in hippocampus, astrogliosis (Kawamata et al., 1998), Aβ aggregates, ↑Tau phosphorylation (Takeda, 2009), ↑microglia, dysfunctional mitochondria (Kawamata et al., 1998), DA neuron degeneration in SN, noradrenergic neuron degeneration in LC (Takeda et al., 1981) Decreased anxiety-like behaviour, disrupted sleep cycle (Miyamoto, 1997), impaired memory and learning (Takeda et al., 1981) Ageing  (Takeda et al., 1981) DNA damage (Shimada et al., 2002), cortical atrophy (Shimada et al., 1994), ↑microglial activation, ↑pro-inflammatory microglial signalling (Kumagai et al., 2007), ↑ubiquitinated inclusions ( Shimada et al., 2008), ↓spine density and synapse density (Shimada et al., 2006) Depressive behaviour, emotional dysfunction (Miyamoto, 1997) SAMP8-APP/PS1 (Lok et al., 2013) 6  (Toth, 2018).Further, the portion of the cohort that remains is arguably biased.In addition, long-term housing of rodents can often be prohibitively expensive and may be limited by potential space constraints of animal holding facilities.Additionally, when animals are housed longterm under so-called standard conditions, it can also lead to chronic stress, behavioural abnormalities and increased risk of health conditions (Cait et al., 2022;Pinelli et al., 2017), requiring alterations to conventional housing protocols when maintaining animals for months at a time.
To counter these practical and ethical difficulties, inbred and genetic models of accelerated ageing have been developed and are discussed below as alternatives to naturally ageing animals in research.Such models, particularly when crossed with well-established genetic models of ND, such as the APP/PS1 mouse model of AD, may allow for accelerated investigations, overcoming the practical challenges associated with long-term rodent investigations, while also modelling disease pathophysiology on a more translationally relevant aged phenotype background.
Mice are the most common model of accelerated ageing, due to their well characterised ageing phenotype and genome (Cai et al., 2022).The senescence accelerated prone mouse (SAMP) model is widely used and was created by selectively inbreeding mice from the AKR/J strain that showed spontaneous age-related changes between 2 and 12mths, such as shortened lifespan and decreased activity, as well as physical features of ageing, including alopecia, cataracts and musculoskeletal changes (Takeda et al., 1981).One strength of SAMP mice is that, as opposed to genetically modified models, acquisition of the aged phenotype is spontaneous and not induced by manipulation of a specific gene or pathway.As biological ageing has no known single cause, inducing models via specific pathways may over-simplify the complexity of natural ageing.In this sense, the SAMP mouse model may be more reflective of the heterogenous nature of human ageing.
A number of sub-strains of the SAMP mouse were generated displaying different phenotypes, two of which (SAMP8 and SAMP10) are particularly relevant to ND (reviewed: (Takeda, 2009)).SAMP8 is widely utilised in AD research, as these mice display both the cognitive deficits seen in AD and AD-specific pathology.Behaviourally, SAMP8 mice have deficits in learning and memory and show decreased anxietylike behaviours and disrupted sleep cycles (Miyamoto et al., 1992).Summarised in Table 2, the brains of these mice show age-related changes (Takeda et al., 1981), as well as amyloid beta aggregates and tau hyperphosphorylation (Takemura et al., 1993).Additionally, SAMP8 mice have shown evidence of a dysregulated immune response, which, as discussed above, has been linked to both ageing and neurodegeneration (Molina-Martínez et al., 2020).Thus, the SAMP8 mouse combines ageing with the presence of relevant neurodegenerative brain pathology (del Valle et al., 2010) and has been used to test the therapeutic potential of multiple agents, including lithium (Malerba et al., 2021;Toricelli et al., 2021), fatty acids (Vela et al., 2019), acupuncture (Wang et al., 2018;Zhang et al., 2013) and both natural (Igarashi et al., 2022;Kim et al., 2023;Shih et al., 2010) and conventional (Lian et al., 2021) medicines.In addition, SAMP8 mice have also furthered our understanding of the role that an array of factors may play in the pathophysiology of AD, including protein chaperones (Akbor et al., 2021b), protein (Liu et al., 2010) and histone deacetylases (Fontán-Lozano et al., 2008) and candidate AD genes, such as SLC24A4 polymorphism (Akbor et al., 2021a).
Interestingly, combining the SAMP8 mouse strain with a mouse model of AD may better represent both the natural ageing changes occurring in these mice, as well as disease pathology.In line with this, the SAMP8 mouse was crossed with the widely used APP/PS1 model of AD (SAMP8-APP/PS1) by Lok and colleagues in 2013 (Lok et al., 2013).The APP/PS1 transgenic model mimics amyloid pathology in the brain and the combined SAMP8-APP/PS1 model demonstrates earlier presentation of pathology, with Aβ immunoreactivity manifesting at 6mths, compared to 9-mths in control mice.Further, an increased number of Aβ deposits were present in the hippocampus in SAMP8-APP/PS1 mice at 9-mths.Acquisition of the AD behavioural phenotype was also more pronounced in SAMP8-APP/PS1 mice, with more severe learning and memory deficits and decreased anxiety at 9-mths of age (Lok et al., 2013).While this suggests that it is feasible to generate a combined model to investigate AD in the context of an aged phenotype, outside of this initial report a decade ago, this model has not been used in further studies and no other combined models have been produced, representing a potential underutilised opportunity to more quickly and accurately model AD.Additionally, despite the presence of relevant pathology, including degeneration of DA neurons in the substantia nigra and noradrenergic neurons in the locus coeruleus in SAMP mice (Karasawa et al., 1997), SAMP8 mice are rarely used for PD research.
In contrast to SAMP8 mice, SAMP10 mice display primarily agerelated changes in the brain and lack the Aβ deposition and tau hyperphosphorylation seen in SAMP8 (see Table 2).There are also behavioural differences between the two strains.While SAMP8 mice show a pronounced decrease in anxiety-like behaviour, SAMP10 mice display increased emotional dysfunction and depressive behaviour (Shimada and Hasegawa-Ishii, 2011).This is thought to be due to a number of changes in the dopamine and serotonin systems of SAMP10 mice, with changes in DA and serotonin receptor sensitivity (Onodera et al., 2000) and altered DA metabolism present in these mice (Takahashi et al., 2004).Interestingly, however, despite these observed changes in the dopaminergic pathway, SAMP10 mice are also underutilised as a model in PD research.

Genetic mouse models of ageing
In contrast to the spontaneous ageing models, genetic models, which involve knockout, mutation, or overexpression of particular genes, are widespread methods in many areas of research, including both ageing and ND.In the context of ageing research specifically, these models involve manipulating a specific pathway or factor to induce premature ageing, which allows for the study of how this pathway/factor contributes to the neurobiological process of natural ageing.An understanding of this is critical for understanding the mechanisms of normal ageing, which lays the foundation for how this may be further altered in the context of age-related ND.Mouse models of the aged CNS have been reviewed previously (Heng et al., 2017), including transgenic models based on genetic changes in progeroid syndromes and specific agerelated pathways, including IL-10 and SIRT6.While multiple genetic models of ageing are available, Sharma et al., (2018) recommended the use of progeroid mouse models particularly to study AD, in order to combine the rapid ageing seen in the progeroid models with the disease phenotype induced by AD models (Sharma et al., 2018).Progeroid mouse models are based on accelerated ageing diseases with a known genetic cause, such as the LMNA gene in Hutchinson Gilford syndrome (HGS), which results in a partially processed pre-lamin A protein, termed progerin (Inesta-Vaquera et al., 2022), and the WRN gene in Werner syndrome (WS), which causes dysregulation in DNA repair systems (Massip et al., 2006).These progeroid syndromes are characterised by early onset of ageing and age-related disease (Ding and Shen, 2008).HGS mice with a mutation in the LMNA gene (LMNA G609G/G609G ) display ageing at three wks, including increased cancer incidence, infertility and curvature of the spine (Osorio et al., 2011).Importantly, however, the brains of these mice do not demonstrate neurodegenerative change, with neurons appearing to not be sensitive to increased progerin expression (Baek et al., 2015).In line with this, HGS mice do not display deficits on tests of reference or spatial memory (Baek et al., 2015) and individuals with HGS do not exhibit deficits in mental or intellectual abilities.This makes the HGS mouse model itself unsuitable for studying ND; however, it may have utility for crossing with established genetic models of ND (e.g., the APP/PS1 model of AD) to produce ND changes in combination with an accelerated aged phenotype.
In contrast, WS mice, which are deficient in part of the helicase domain of the WRN protein (WRN Δhel / Δhel ), show an ageing phenotype with onset at 5 mths, including insulin resistance, increased ROS, DNA damage and increased cancer incidence (Massip et al., 2006).Additionally, both motor (loss of activity and coordination) and cognitive (deficits in spatial and social novelty memory) changes, concomitant with signs of cellular ageing and enhanced oxidative stress and neuroinflammation, have been described in this model (Hui et al., 2018) (Table 2).This suggests that these mice may have utility in their own right for the study of ND, with studies needed to examine markers associated with neurodegenerative change (e.g., tau hyperphosphorylation).It is also important to note, however, that this potentially limits the ability to cross WRN mice with established genetic models of ND (e.g., the APP/PS1 model of AD) to produce mixed models, as the presence of such behavioural impairments may confound interpretation of results.
This has not, however, prevented the generation of a mouse model with a combined complete KO of both WRN and telomerase RNA component (Terc), a ribonucleoprotein polymerase that maintains telomere length, (Terc − / − WRN − / − ) (Chang et al., 2004).Terc − / − mice display accelerated ageing at just 4-8 mths old, as well as upregulation of SASP related genes, increased oxidative stress response and apoptotic markers in the brain (Yang et al., 2022).To date, Terc − / − WRN − / − mice have not been assessed for neurological or behavioural effects, but since both Terc − / − (Khan et al., 2015) and WRN modified (WRN Δhel / Δhel ) (Hui et al., 2018) mice have neurological impairments and cellular brain changes (Yang et al., 2022), it is reasonable to predict that Terc − / − WRN − / − mice may display exacerbated behavioural change and a greater extent of age-related pathophysiological change, compared to either model alone.This may make this model particularly useful for ND research, although this has yet to be evaluated.Nevertheless, Terc knockout mice have previously been utilised to study PD.Scheffold and colleagues (2016) crossed C57/BL6/J Terc − / − mice with C57/BL6/J mice expressing the human mutant α-synuclein (A3oP) to model PD in the context of telomere shortening (Scheffold et al., 2016).Results demonstrated that telomere shortening accelerated the formation of α-synuclein aggregates and impaired responses of brainstem microglia (Scheffold et al., 2016).This study provided evidence that age-related changes in the brain, such as telomere shortening, impact the progression of degenerative pathology in PD, further highlighting the importance of accounting for the effects of ageing when modelling ND.
In addition to progeroid and telomere shortening, genetic models based on inflammatory and metabolic theories of ageing have also been utilised, including IL-10 (Walston et al., 2008) and SIRT6 deficiency (Kaluski et al., 2017).Mice deficient in the anti-inflammatory cytokine IL-10 (IL-10 − / − ) had previously been established as a model of human frailty, showing alopecia and decreased muscle strength at 11mths (Walston et al., 2008), low-grade activation of inflammatory signalling pathways, including IL-6, TNF-α and IL-1β, at 15mths (Ko et al., 2012), and increased neuroinflammation and Tau phosphorylation at 6mths old following an inflammatory stimulus (Weston et al., 2021).While IL-10 − / − mice display brain changes similar to those in AD (Richwine et al., 2009), however, they are currently not widely used in ND research.
Similarly, SIRT6, a member of the Sirtuin family of signalling proteins, is important for maintaining normal DNA repair (Mostoslavsky et al., 2006) and has been the basis for modelling ageing in C57BL/6 mice (Mostoslavsky et al., 2006).SIRT6 − / − mice demonstrated genomic instability, altered metabolic homeostasis and immune cell generation and decreased size (Mostoslavsky et al., 2006).While these changes demonstrate key elements of the mammalian aged phenotype discussed above, SIRT6 − / − mice develop degenerative progeroid pathology and do not survive beyond 4wks (Mostoslavsky et al., 2006), limiting the ability of researchers to conduct experiments.However, mice with a neuronal specific SIRT6 deficiency survive longer and demonstrate ageing, measured by DNA damage and cell death, at 4mths of age (Kaluski et al., 2017) (Table 2).Considering the deficits in nonassociative learning and increases in locomotor activity in this model are similar to those in AD models, and that SIRT6 expression in the brain is reduced in AD patients compared to aged controls (Kaluski et al., 2017), SIRT6 − / − mice may be useful for modelling AD in the context of an accelerated aged phenotype.Of note, SIRT6 − / − embryonic stem cells have previously been used as an in vitro model of age-related cellular changes (Mostoslavsky et al., 2006) (Table 1).
Of potential relevance, a new genetic rodent model of ageing was reported in January 2023, the inducible changes to the epigenome (ICE) mouse.This mouse was created by tamoxifen-controlled expression of the endonuclease I-PpoI at 20 canonical sites in the mouse genome, causing DSBs without mutations in the genome (Yang et al., 2023).When I-Ppol was activated at 4-6mths, an aged phenotype was observed by 10-12mths.This featured muscle loss, alopecia, loss of skin pigment and premature frailty, similar to that of 24mo controls (Yang et al., 2023).Additionally, higher levels of activated glial cells in the hippocampus suggest potential applications in modelling the brain changes seen in NDs.The researchers additionally investigated the ability to reverse these changes by treatment with Yamanaka factors, which are used to differentiate somatic cells to iPSCs.Ageing changes observed in ICE mice were able to be reversed in vitro, demonstrating the potential usefulness of the Yamanaka factors beyond iPSC differentiation.It is important to note, however, that there have been a number of major issues identified with the design of this study (Timmons and Brenner, 2023), most notably that genotoxicity due to the expression of I-Ppol appears to potentially play a role in development of the ageing phenotype in ICE mice through p53 signalling (Kim et al., 2016), which has not been accounted for in controls of this study.However, while the specific mechanisms driving ageing in this model are yet to be fully validated, this method presents a potentially significant contribution to the research field by providing an inducible rodent model of ageing that could be crossed with established genetic models of ND disease.Regardless, further characterisation of the neurobiological changes in these mice needs to be undertaken to understand the utility of this model in age-related research, as well as for CNS research more generally.
While potentially promising, particularly with regard to their ability to overcome the practical challenges associated with the long-term investigations needed to investigate spontaneous ageing, it is important to note that accelerated ageing models are not without their own limitations and challenges.Namely, there are important questions about how closely these models mirror the changes associated with natural ageing.In line with this, it has been observed that the histopathologic lesions associated with these accelerated ageing models often don't overlap completely with those seen in natural ageing (Gurkar and Niedernhofer, 2015), and that accelerated ageing models often exhibit features not typical of normal ageing (Kõks et al., 2016).Similarly, none of the progeria models reviewed here are able to mimic all of the complex aspects associated with the aged phenotype in mammals due to the fact that they were created to specifically model the changes seen in certain diseases (Gurkar and Niedernhofer, 2015).It is also important to note that the morphologic changes observed in these models are often segmental in nature, meaning that multiple different models would be required in order to fully explore the complex changes that contribute to the aged phenotype within a particular tissue or organ system (Harkema et al., 2016).

Considering age in NDs moving forward-recommendations for researchers
A major gap in ND research currently is the lack of consideration given in current disease models to the typical changes associated with the aged phenotype that occur alongside disease-specific pathology.Fig. 2 outlines a proposed research pipeline to integrate ageing into ND research both in vitro and in vivo using currently available models.First, the aged phenotype is induced in either a relevant cell or animal model (Fig. 2 step1), in order to simulate the complex cellular environment present in the aged individual.Once this has been established, a disease model can then be induced on this background using an inflammatory (Hoban et al., 2013;Mutemberezi et al., 2018), chemical (Sherer et al., 2003) or protein stimulus in vitro (Ross et al., 2020) and, additionally, genetic (Flood et al., 2002) or neurotoxic methods (Ungerstedt, 1968) in vivo (Fig. 2 step 2).This more translationally relevant model can then be used to undertake the desired investigations (Fig. 2 step 3).The incorporation of this pipeline would benefit both the understanding of disease pathogenesis, by providing insight into the contribution of biological ageing to NDs, and additionally would assist in the development of disease treatments by providing more accurate screening of potential therapies.Major drawbacks of this pipeline include not only the time needed to establish the combined ageing and ND model, but that the complexity of the generated model introduces the need for extra validation and requires a more multifaceted interpretation of results.To address the issue of time, utilising the accelerated in vivo ageing models reviewed here would alleviate some of this cost compared to using naturally ageing animals; however, none of these models individually can capture all the age-related changes described above.Establishing such models in vitro is far less costly and time consuming; for example, inducing replicative senescence in primary cells, which have a limited lifespan, would be considerably faster and cheaper than housing mice for 18 months or longer.Modelling the complex processes seen in both ageing and NDs presents significant challenges for researchers both in vitro and in vivo.Below, we present a discussion of these challenges, and in Table 3, we offer some recommendations to consider when designing both in vitro and in vivo ND studies.
As reviewed above, despite the array of techniques used to induce and quantify ageing in vitro and in vivo, there is yet untapped potential for incorporating the aged phenotype in preclinical ND research.First, when it comes to in vitro work, current models are often over-simplified and fail to recapitulate the complex 3D interactions amongst different types of brain cells in the tetrapartite synapse.Moving forward, use of co-culture and organoid models kept for longer periods of time in culture (i.e.27 day old co-cultures (Bigagli et al., 2016) or 45 day old brain organoids (Ao et al., 2022) displaying ageing markers) can be drawn upon, in combination with iN technology, to produce sophisticated "aged" in vitro models that allow for investigation of disease mechanisms or treatment efficacy in more representative models.In order to validate such models, use of a standardised "aged phenotype" marker panel, inspired by the 'AgeScore' panel of Hartmann and colleagues (2023) (Hartmann et al., 2023), should be developed and consistently applied across studies in order to improve reproducibility and allow for more ready comparison between labs.By utilising cells which meet a minimum criterion on this panel, or a minimum 'AgeScore', ageing can be equally accounted for across experiments and between lab groups.While this panel needs further validation across cell types, we suggest that for a culture to be considered 'aged,' cells must show changes in major cellular ageing markers discussed as part of the aged cellular phenotype above and identified as a hallmark of ageing.For ND models, the acquisition of the SASP is particularly important, as chronic inflammation has well established links to neurodegenerative pathology (Guzman-Martinez et al., 2019) and ageing (Franceschi and Campisi, 2014) and should serve as a major marker of ageing in any potential aged ND model.This could serve a similar function to the ARRIVE guidelines (Animal Research: Reporting of In Vivo Experiments), which have sought to improve reporting of research in animal studies (Percie du Sert et al., 2020), and could also be used for validation of in vivo models.The generation of such well-validated, sophisticated "aged" in vitro models will provide a powerful tool on which to model ND and could be used in combination with exciting advances in bioengineering technologies, such as microfluidic devices (Fernandes et al., 2016) (e.g.brain-on chip) and 3D tissue engineering (Lam et al., 2019) (e.g.extracellular matrix decellularization) to take our understanding of disease mechanisms and drug development to new heights.
From an in vivo perspective, despite recent improvements in animal models of ageing, there is an overall lack of consideration of age in ND study design.In a survey of 297 researchers on the use of rodents in biomedical research conducted by the National Centre for the Replacement, Refinement and Reduction of Animals in Research (NC3R), the most common age of use is 8-12 weeks (Jackson et al., 2017) (with 6-20 weeks being reported as "adult" animals).This is problematic, as this is a stage in which many developmental processes are still ongoing, including, importantly, both development of the immune system (Holladay and Smialowicz, 2000) and brain maturation (Downes and Mullins, 2014) and myelination (Fu et al., 2013).Further, as discussed extensively above, the physiological changes that occur with ageing differ significantly in young versus old animals.Concerningly, however, only ten percent of respondents to the survey reported using rodents over 16 weeks of age, with just 24 % of these in in rats (Jackson et al., 2017).This is regardless of whether age is a significant risk factor for the disease being modelled (e.g.NDs) and in spite of a growing number of studies showing the moderating effect of age on disease presentation and pathophysiology (e.g.older rats have increased susceptibility (Yager and Thornhill, 1997) to and take longer to recover from (Brown et al., 2003) ischaemic insults).
To improve upon this, animal age must be carefully considered when designing studies.For NDs, this means planning for either the long-term use of animals (i.e.18-months or longer) or the use of an accelerated ageing model, such as those reviewed above.In selecting an accelerated ageing model, particularly those based on progeroid syndromes, researchers must consider the different pathophysiological changes that occur in each model, when these changes occur and how this may impact their interpretation of findings (e.g.cognitive impairment seen in WRN mice may confound behavioural results in a mixed model of AD).Due to the difficulties in interpretation of results and the limited hallmarks of ageing displayed by any one accelerated model, using naturally aged animals is the ideal base model to use in step 1 (Fig. 2).However, when using naturally aged animals, there are important financial and practical difficulties associated with maintaining rodent colonies long term.For example, a comparison by NC3R of the cost of obtaining animals at six versus 12-weeks of age from a major commercial supplier in the US and UK demonstrated that 12-week animals cost at least 50 % more per animal, and sometimes as much as 100 % more (Jackson et al., 2017), and this is without accounting for increases in agistment costs associated with ageing rodents in-house.Improvements in the availability and accessibility of aged animals, such as better commercial availability, local/national networks for aged animal sharing or improved banking of tissues or findings into biobanks/data repositories for secondary analyses, would help to offset the significant costs, challenges and, most importantly, variability associated with individual research groups housing animals for extended periods of time within their own housing facilities, in order to obtain aged cohorts.Some efforts have already been made in this direction via the Shared Ageing Research Models (ShARM) initiative (Duran et al., 2013) and the Aging Rodent Colonies project at the National Institute on Aging (Jackson et al., 2017).Additionally, researchers must work together to advocate to government and philanthropic bodies to provide adequate funding to allow for the increased costs associated with the use of aged animals within grant applications, where appropriately justified.
The field would also benefit from the development of guidelines for animal care and use specifically tailored towards older animals, given the unique challenges associated with these cohorts.For example, rats have been shown to develop multiple diseases and alterations in organ function in conjunction with ageing, including renal disease (Roncal-Jimenez et al., 2016), respiratory disease/infections (Carthew et al., 1989), increased tumour presentation (Sandusky et al., 1988) and atrophy of lumbar/hindlimb muscles (Ibebunjo et al., 2013).Further, such age-related changes may be strain specific, which also must be accounted for when working with aged animals.In support of this, C57BL/6J mice have missense mutation in the nicotinamide nucleotide transhydrogenase gene, which leads to impaired glucose tolerance (Toye et al., 2005) and age-related declines in mitochondrial function (Ghosh et al., 2014).Researchers would thus benefit from sharing experiences of working with aged cohorts and from working closely with laboratory animal servicers and veterinary staff at the time of experimental design, to anticipate challenges that may arise and appropriately plan how to manage these.Importantly, there are also several extraneous factors not discussed in this review which can influence ageing markers and potentially contribute to heterogeneity in human populations, including stress (Yegorov et al., 2020), exercise (Carapeto and Aguayo-Mazzucato, 2021) and diet (Longo et al., 2021); however, these factors need to be considered and controlled for in the study design phase, keeping in mind that the needs of older animals may not be equivalent to those of younger animals.
Finally, in accordance with 8a of the Essential 10 of the ARRIVE guidelines, increased emphasis needs to be placed on appropriate reporting of animal age.This includes both summary statistics (mean +/− SD, as well as range) for each experimental group and, importantly, values for individual animals in the supplementary materials (Percie du Sert et al., 2020).This is critical, given that animal age is often poorly reported, or even unreported, in preclinical studies, with poor justifications offered for age choices (e.g.monetary costs, supply/availability and maintaining comparability with historical data) (Jackson et al., 2017).This leads to wide discrepancy in the age of animals being used by different labs, with an investigation of this by NC3R showing that, within the same model paradigm, the age of the animal used could vary over a range of up to 20 weeks (Jackson et al., 2017).Such inconsistencies in age choice and reporting can lead to variability between animals within a given study, as well as between studies using the same Table 3 Summary of recommendations for improving the modelling of age both in vitro and in vivo.

In vitro recommendations
Develop more sophisticated in vitro models as a background on which to model NDs using advances in co-culture and organoid technology Validate "aged" models using a standard marker panel and use this to develop reporting guidelines (in line with the ARRIVE guidelines for in vivo research Harness advances from bioengineering (e.g.brain on a chip and extracellular metric decellularization) to use in combination with aged in vitro models in order to allow for deeper investigation of disease mechanisms

In vivo recommendations
When designing studies, utilise models which include changes relevant to the disease of interest (e.g.pathological protein deposition in ND studies) Advocate for improvements in the availability and accessibility of aged animals, including better commercial availability, local/national networks for aged animal sharing or improved banking of tissues or findings into biobanks and data repositories for secondary analyses Educate governmental and philanthropic bodies about the importance of incorporating aged animals into research and the need for increased funding to allow for this, where justified Develop guidelines for animal care and use specifically tailored towards older animals that also account for strain-specific differences Work closely with laboratory animal services and veterinary staff when designing experiments, in order to anticipate challenges that may arise and develop strategies to manage these.Adhere to the ARRIVE guidelines for the appropriate reporting of animal age in peer-reviewed literature, reporting both summary statistics for each experimental group and, where possible, age data for individual animals in the supplementary data Utilise tools developed to determine human equivalent age when selecting the age of animals to be used in a given study, in order to facilitate the development of a field-specific consensus about age of animals to be used in a given model moving forward L.M. Carr et al.
model from different laboratories, negatively impacting data quality and study outcomes (Jackson et al., 2017).Additionally, there needs to be greater consensus in the field about what constitutes specific age ranges of interest in a given species (e.g.younger adult versus middle-age versus older adult) and how this relates to human ageing.
To assist with this, as discussed above, recommendations have been put forward by the Jackson Laboratory, proposing three life stages in mice: mature adult from 3-6 months of age (~20-30 human years), middle age from 10-15 months of age (~38-47 human years) and old age from 18-24 months (~56-69 human years) (Jackson Laboratory, 2020).Researchers can use these as a guide when designing studies which utilise aged mice; however, potential strain and species differences also need to be considered.In order to facilitate this, additional tools have been developed to help determine equivalent human age in mice (Geifman and Rubin, 2013) and rats (Quinn, 2005), as well as other species (Clancy et al., 2007).Ideally, researchers should aim to utilise animals at the higher end of the defined "old age" range (e.g.24-months of age in C57BL/6J mice), although potential health concerns for animals of this age, as well as practical and financial concerns associated with long-term housing, also need to be considered.

Conclusions
Overall, there is a critical need to incorporate ageing into studies of neurodegeneration, which is currently often overlooked by researchers, despite age being the most significant risk factor for ND development.
Whilst we highlight that it is critical to include age in ND research to improve translatability, we acknowledge that all models have limitations and no one model can recapitulate all elements of human disease.ND research conducted without specific consideration of age still presents an invaluable contribution, as evidenced by the multiple advances made in the field to date.Nevertheless, moving forward, effort should be made to account for ageing in study design or consider it in the interpretation of results.This is particularly critical given how age may alter not only pathophysiology, but also potentially the response to therapeutic interventions.Therefore, whilst we also recognise the difficulties involved in analysing such convoluted data, the interactions between NDs and ageing make the diseases themselves complex.Thus, a complex model is more reflective of the naturally occurring disease environment.Here, we have summarised what constitutes the aged phenotype in humans and critically evaluated several major methods to induce ageing in a laboratory setting, with a particular focus on the significant underutilisation of these models to date in ND research.Emerging in vitro technologies could help to improve this, broadening the landscape of available models and allowing for more sophisticated investigations of disease mechanisms and therapeutic response in the context of a complex environment more reflective of the CNS.Similarly, the advent of in vivo models of ageing that are both inducible and reversible may help us to probe the contribution of ageing to the emergence of the NDphenotype in a way never before possible.In spite of these advancements, however, it is critical that ND researchers prioritise including age as a critical component of their models moving forward.With this in mind, and in the spirit of overcoming current practical challenges, we have offered here specific recommendations for improving the incorporation of the aged phenotype into both in vitro and in vivo experimental models.Taken together, it is hoped that more widespread use of these models, in conjunction with our recommendations, will improve the quality and reproducibility of the data produced, ultimately helping to ensure that results can translate to effective treatments for neurodegeneration.

Fig. 2 .
Fig. 2. Pipeline for the incorporation of age into ND research models.Step 1: Either in vitro or in vivo, studies need to induce the senescent phenotype in cells or ageing in animals.Step 2: Once ageing has been induced, disease can then be modelled within those animals/cultures.Types of models to induce disease include genetic, inflammatory, seeding, or neurotoxic models Step 3: biological responses of cells/tissue and behavioural responses of animals can then be tested to develop treatments and understand disease pathology.

Table 1
Strengths and weaknesses of in vitro methods to model CNS ageing.

Table 2
Strengths and weaknesses of in vivo methods to model CNS ageing.