1 Introduction

1Neurodegenerative diseases (NDs) affect millions of people worldwide, with ageing being a primary risk factor. These disorders, characterised by progressive neuronal loss, include Alzheimer’s disease (AD), Parkinson’s disease (PD), Huntington’s disease (HD), and amyotrophic lateral sclerosis (ALS), each with distinct yet often overlapping pathophysiological features such as protein aggregation, neuroinflammation, and synaptic dysfunction. Despite significant research efforts, effective disease-modifying therapies remain elusive.

2Rodent models have been instrumental in elucidating ND mechanisms but fail to recapitulate the human brain's complexity fully (Wallis, 2011; Roelfsema & Treue, 2014; Strange et al., 2014). Differences in cortical organisation, neurotransmitter systems, and cognitive abilities limit the direct translation of rodent findings to human clinical trials. As a result, non-human primates (NHPs) have emerged as valuable models due to their closer phylogenetic relationship to humans (Rhesus Macaque Genome et al., 2007). This paper explores the relevance of NHP models in ND research, their contributions to understanding disease mechanisms, and the challenges and future directions of their use.

3NHPs, including macaques, marmosets and lemurs, share significant genetic and neuroanatomical similarities with humans. These similarities make them ideal for studying neurodegenerative disorders, particularly those affecting higher-order cognitive functions, motor control, and complex behaviours. Unlike rodents, NHPs possess a highly developed prefrontal cortex, a key region in cognitive decline in disorders like AD (especially the old-world NHPs). Furthermore, the spontaneous occurrence of amyloid plaques and tau pathology in aged NHPs strengthens their utility as AD models (Isidro, 2024), a feature likely related to the evidence that macaques and humans express 3R and 4Rtau isoforms, both aggregated in AD. In contrast, adult wild-type rodents only express 4R tau. They also exhibit a dopaminergic system comparable to humans, making them particularly relevant for PD research. Contrary to rodents, NHPs exhibit a small proportion of monosynaptic corticospinal projections on motoneurons pertinent to ALS (Lemon et al., 2004). NHPs facilitate identifying biomarkers that predict disease progression, as they exhibit cerebrospinal fluid (CSF) and blood biomarkers similar to humans. Brain size makes them compatible with clinically relevant imaging techniques, such as magnetic resonance imaging or positron emission tomography, for the validation of ligands. Longitudinal studies in NHPs allow researchers to track disease evolution in a manner that closely resembles human conditions, with endpoints that are the same as those used in the clinic. For instance, cognitive disabilities could be assessed in NHP models of NDs and in patients using the very same neuropsychological tests on a CANTAB (Cambridge Neuropsychological Test Automated Battery) battery (delayed-matching to sample and paired associates learning for AD, attention and reaction time tasks and intra-extra dimensional set shift for PD).

4NHPs are crucial in preclinical therapeutic evaluations, particularly for small molecules, gene therapies, cell therapies, neurotrophic factors, and biomedical solutions such as deep brain stimulation (DBS) or brain-computer interfaces. Their use in pharmacokinetic and pharmacodynamic studies ensures better translation of therapies to human trials. The perfect illustration of the use of these NHP translational studies is those which defined basal ganglia circuitry (Mitchell et al., 1989; Wichmann et al., 1994; Wichmann et al., 1994; Wichmann & DeLong, 2003), soon followed by the demonstration that lesioning the subthalamic nucleus alleviated PD-like cardinal symptoms in MPTP NHPs (Bergman et al., 1990; Aziz et al., 1991), eventually leading to the discovery of the profound symptomatic relief afforded by subthalamic nucleus DBS in MPTP NHPs (Benazzouz et al., 1993) followed a few months later by PD patients (Pollak et al., 1993; Benabid et al., 1994), a procedure giving life-altering benefit to hundreds of thousands of PD sufferers.

5NHP models of NDs can be categorised into three main types: phenocopic, pathogenic, and aetiologic models (Bezard & Przedborski, 2011). Phenocopic models aim at replicating the spectrum of symptoms using various strategies. Pathogenic models attempt to replicate the observed pathology in NDs, i.e. multi-systemic neurodegeneration, presence of the anatomopathological landmarks of the considered ND (amyloid plaques, Lewy bodies, etc.) and central/peripheral neuroinflammation. Aetiologic models focus on studying the consequences of genetic manipulation of NDs-related loci. While phenocopic models help validate therapeutic strategies to alleviate a given symptom, pathocopic and aetiologic models, which mostly rely on biological endpoints, support the development of the much-awaited ND course-modifying therapies.

2 NHP Models of Specific Neurodegenerative Diseases

2.1 Alzheimer’s Disease (AD)

6AD is pathologically characterised by amyloid-beta (Aβ) plaque deposition, tau neurofibrillary tangles, synaptic loss, and neuroinflammation. Intended to be phenocopies of AD and mild cognitive impairment (MCI), as well as in evaluating potential therapeutic interventions, some pharmacological models aim to replicate key features of human cognitive decline, including memory deficits. Ageing on its own induces an MCI-like state (Buccafusco, 2006; Languille et al., 2012; Schneider et al., 2013; Castonguay et al., 2018; Frye et al., 2022). Scopolamine, an anticholinergic agent, is commonly used to transiently impair memory in NHPs by disrupting cholinergic neurotransmission. While effective in inducing memory deficits (Buccafusco, 2009), this model is considered nonspecific, as it does not fully replicate the complexity of MCI or AD pathology (Pepeu, 2004). Intracerebral injection of amyloid-β (Aβ) oligomers in NHPs has been shown to induce AD-like pathology, including synaptic loss, tau hyperphosphorylation, and cognitive deficits (Beckman et al., 2019). This approach provides a more accurate representation of human AD pathology but has rarely been used in pharmacological settings. In summary, pharmacologically induced cognitive impairment models in NHPs are essential in neuroscience research, offering a closer approximation to human cognitive disorders than rodent models. Continued refinement of these models is crucial for advancing our understanding of AD and MCI and for developing effective treatments.

7The recent boom in pathocopic models of AD in NHPs further illustrates the pathological limitations of the pharmacological impairments. Still, it stresses how challenging it is to produce AD-like cognitive impairments based on a pathological trigger. Indeed, most of the below-reported strategies, while delivering amazing, although restricted and limited, pathologies, are rarely accompanied by reproducible cognitive deficits.

8While transgenic mouse models have provided insights into AD pathology, they do not fully replicate human-like tau aggregation or late-stage neurodegeneration (Qian et al., 2024). In NHPs, naturally occurring Aβ and tau pathology have been observed in aged individuals, making them valuable for studying spontaneous AD progression. Great apes, macaques and lemurs develop amyloid plaques with age, in a similar distribution pattern to that observed in humans, but the development of end‐stage neurofibrillary tangles (NFT) and the extent of tau hyperphosphorylation in NHPs is still controversial (Languille et al., 2012; Beckman et al., 2021; Souder et al., 2021). A review of the literature suggests, however, that while amyloid pathology follows an allometric pattern vis-à-vis the lifespan in NHPs compared to humans, the tau pathology would evolve chronometrically, explaining why full-blown dementia and subserving pathology is not observed in very old NHPs.

9As reported above, intracerebral injection of Aβ oligomers in NHPs induces AD-like pathology, including synaptic loss, tau hyperphosphorylation, and cognitive deficits (Beckman et al., 2019). However, because of this isometric Tau behaviour with regard to ageing, a fantastic boom around Tau pathology in NHPs has recently emerged. The viral-vector-mediated overexpression of wild-type or mutated tau into the NHP entorhinal cortex (Beckman et al., 2021; Beckman et al., 2024) or hippocampus (Jiang et al., 2024) leads to tau pathology. Pathology was reported to span the entorhinal cortex connectome, including the hippocampal formation. Comparably, macaques stereotactically injected with patient-derived aggregated tau extracts faithfully reproduce the tau pathology and spread off (for all tauopathies, actually) (Darricau et al., 2023; Darricau et al., 2024; Darricau et al., 2025). Initial attempts to induce a full spectrum pathology associated with some cognitive impairment had been successfully obtained in the lemur by exposing them in the parietal (Gary et al., 2019) or cingulate (Lam et al., 2021) cortex to patient-derived brain homogenates (crude, non-purified for a specific protein). Such patient-derived tau extracts, when originating from a different tauopathy, e.g. the progressive supranuclear palsy, lead to a progressive supranuclear palsy-like pathology (and behavioural manifestations) supporting the concept of diseases-specific polymorphs (Darricau et al., 2023). There are other modelling strategies being developed, although still very preliminary, such as the rodent-derived intracerebroventricular streptozotocin model (Yeo et al., 2015) or the ecological spontaneous type 2 diabetic cynomolgus monkey as a preclinical MCI model (Huang et al., 2024).

10Genetically modified NHP models, such as marmosets (Yoshimatsu et al., 2022; Kohri et al., 2023) or macaques (Seita et al., 2020; Tu et al., 2023) expressing human APP mutations or mutated tau, have also been developed to mimic early-stage AD pathology with mixed outcomes (Pan et al., 2024). As for the other conditions, various gene editing strategies have been used, making almost every single paper a case report. The field is still far from being able to use these transgenic animals in large-scale translational therapeutic testing conditions.

2.2 Parkinson’s Disease (PD)

11PD is primarily driven by dopaminergic neuron loss in the substantia nigra and the accumulation of a-synuclein aggregates (Lewy bodies). Rodent models fail to accurately reproduce the selective vulnerability of midbrain dopaminergic neurons seen in humans, notably because they do not accumulate neuromelanin as they age.

12MPTP-induced NHP models have been extensively used to study PD since MPTP effect discovery in a small cohort of Californian addicts. MPTP, a neurotoxin, destroys dopaminergic neurons, leading to motor and non-motor symptoms akin to human PD. The various regimens of MPTP intoxication fall under the denomination of phenotypic models (Bezard & Przedborski, 2011; Bezard et al., 2025). These models have been instrumental in developing L-DOPA therapies, adjunct pharmacological therapies for controlling L-DOPA-induced dyskinesia, DBS, gene therapies targeting dopaminergic pathways and, more recently, brain-computer interfaces for restoring gait. They have more recently helped characterise PD-like non-motor symptoms such as cognitive impairments, freezing of gait, pain, and sleep disturbances (Bezard et al., 2025).

13The report on the genetic association of -synuclein mutation and familial parkinsonism (Polymeropoulos et al., 1997) launched the race to identify genetic causes for PD. The few monogenic (mendelian) forms of PD (SNCA, LRRK2, GBA1, etc...) are accompanied by a large variety of genetic risk factors identified in genome-wide association studies. Viral vector-based aetiologic models have thus been developed with overexpression of wild-type or mutated -synuclein (Kirik et al., 2002; Kirik et al., 2003; Eslamboli et al., 2007; Koprich et al., 2010; Bourdenx et al., 2015) or downregulation of genes such as ATP13A2 (Sikora et al., 2024). Falling into the aetiology category but without a link to genetics is the tyrosinase model. Neuromelanin is primate-specific. Although a key feature of human dopamine neurons and likely involved in oxidative stress, it has only been recently scrutinised. This is an innovative aspect investigated by Vila and co-workers first in rodents (Carballo-Carbajal et al., 2019; Vila et al., 2019) and very recently in NHPs (Chocarro et al., 2023) where they used an adeno-associated virus (AAV) 2/1 encoding the human tyrosinase gene (hTyr). Following stereotactic nigral injection of AAV-hTyr, and after 4 to 8 months, the study revealed a time-dependent accumulation of neuromelanin. This accumulation was found to initiate the pathological misfolding of endogenous α-synuclein in pigmented midbrain dopaminergic cells, with the potential to propagate anterogradely toward the cerebral cortex (Chocarro et al., 2023), making this model the first model to reconcile several aspects of the human disease.

14The seminal work of Braak and colleagues suggesting that LB pathology follows a predictable pattern of progression within the PD brain (Braak et al., 2003), as well as the ‘host-to-graft’ observation (Kordower et al., 2008; Li et al., 2008; Mendez et al., 2008) led to the development of pathogenic models based on injection with -synuclein (the most represented protein component of Lewy bodies) assemblies. Developing such pathogenic prion-like NHP models culminated with several efforts using patient-derived extracts more or less purified for their synuclein content (Recasens et al., 2014; Dehay & Bezard, 2019; Bourdenx et al., 2020; Fayard et al., 2023; Darricau et al., 2025) or recombinant--synuclein-preformed fibrils (PFFs; Shimozawa et al., 2017; Chu et al., 2019; Guo et al., 2021; Sawamura et al., 2022; Fayard et al., 2023). As for Tau, intracerebral injection of brain-derived synuclein-positive extracts originating from other diseases, e.g. multiple system atrophy or dementia with Lewy bodies, led to the development of multiple system atrophy- (Teil et al., 2022) or dementia with Lewy bodies-like (Teil et al., 2023) anatomopathology, respectively. Although the pathology is strikingly similar (and informative) to the human PD pathology/ies, none of these models presents however overt parkinsonism like the MPTP macaque model.

15Due to ethical, economic, and technical challenges, few studies have developed transgenic NHP models of synucleinopathy. Niu et al. created A53T-α-synuclein transgenic monkeys via lentiviral injection into fertilized eggs, leading to five births (Niu et al., 2015). These monkeys showed increased α-synuclein in the SNc, striatum, and cortex, but not its phosphorylated form. By 2.5 years, some exhibited cognitive and behavioural deficits resembling early PD, though no motor symptoms or neurodegeneration were seen via MRI. Later, Yang et al. (2019) used CRISPR-Cas9 to target the PINK1 gene in rhesus embryos, producing several mutant monkeys. Some showed decreased movement, grey matter loss, and neuron degeneration by 1.5–3 years (Yang et al., 2019). Although α-synuclein alterations were not observed, the study demonstrated the potential of CRISPR for PD models in NHPs. These works are elegant demonstrations of the technical feasibility. However, the extremely low production yield precludes the generalisation of the approach.

2.3 Huntington’s Disease (HD)

16HD is a genetic disorder caused by CAG repeat expansions in the huntingtin (HTT) gene, leading to progressive striatal degeneration and motor, cognitive, and psychiatric disturbances. The first models aimed at replicating the massive neuronal loss and atrophy of the striatum encountered in HD. These phenocopic models were merely interested in replicating the symptoms and understanding the cortico-basal ganglia pathophysiology resulting in movement and cognitive manifestations. Very much like for the other basal ganglia-related ND, PD, neurotoxins were first used either by direct stereotaxic administration (ibotenic acid, kainate, quinolinic acid) in marmosets (Kendall et al., 2000), baboons (Hantraye et al., 1990) and macaques or alike (Ferrante et al., 1993; Roitberg et al., 2002; Lavisse et al., 2019). Systemic repeated and slow administration of metabolic toxins (3-nitro propionic acid, malonate) interfering with the complex II of the mitochondrial respiratory chain led to a prototypical degeneration of the dorsolateral putamen and dorsal caudate nucleus, manifesting behaviourally as a hyperkinetic syndrome (Palfi et al., 1996; Palfi et al., 2000). These phenocopic models have been used extensively to develop MRI and PET imaging (Brownell et al., 1994; Flament et al., 2018) and validate cell replacement therapies in HD (Schumacher et al., 1992; Aron Badin et al., 2019; Rosser et al., 2021). Although out of fashion these days, they remain fascinating, especially for their metabolic dimension, with the environmental concerns about the widespread use of succinate dehydrogenase inhibitor biocides/pesticides.

17The most fascinating primate HD model is the transgenic NHP modelling that falls into the aetiologic category. HD has been the testbed for transgenesis in NHP because of the well-known genetic cause and the undisputable behavioural manifestation if anything should have occurred. Transgenic NHP models carrying mutant HTT have been developed, exhibiting motor deficits, brain atrophy, and neuroinflammation similar to human HD (Yang et al., 2008a; Chan et al., 2015). The transgenic HD macaques were obtained by injecting lentivirus into the perivitelline space of fertilised embryos (Yang et al., 2008a). However, although the lentivirus transferred the exogenous gene into host cell genome, the transgene insertion site and copy number were uncontrollable (Yang et al., 2008a), leading to pronounced phenotypic variation among the transgenic offspring. The animals carrying 84 CAG repeats displayed severe phenotypes at an early postnatal stage, in contrast to 84 CAG repeats mice, which show subtle symptoms (Yang et al., 2008b; Farshim & Bates, 2018). The second generation of transgenic HD macaques had a 67-72 CAG repeat length (Chan et al., 2015). The slow appearance of rather subtle symptoms was supported by a lesser pathology severity in these 67-72 CAG repeat NHP than in the 84 CAG original NHPs. These results nevertheless paved the way to great hope. It was possible to obtain a relatively rapid phenotype, confirming that NDs are primate-specific in their expression and exhibit a greater susceptibility than rodents.

18The phenotypic variability and the long time required to obtain a group of comparably affected adult NHPs prevent for envisioning using the transgenic modelling strategy for routine preclinical studies. In light of this inescapable reality, pathocopy models might be of interest through either viral overexpression of mutated huntingtin with a variable number of CAG repeats or intracerebral administration of Huntington patient brain-derived extracts (Gosset et al., 2020; very much like the proteinopathy models of PD, AD, multiple system atrophy, dementia with Lewy bodies, Progressive supranuclear palsy). Lentiviral vectors were first used to overexpress in the macaque striatum the first exon of the HTT protein gene with either a 171HTT-19Q (wild type) or a 171HTT-82Q (mutated; Palfi et al., 2007). Convincing pathology (presence of ubiquitinated HTT inclusions in neurites and cell bodies with an associated striatal neuronal loss) was associated with spontaneous motor abnormalities. More recently, AAV were tested with the same construct (Aron Badin, 2018) or shorter construct (Exon1 - 19Q - wild type or 103Q - mutated) triggering pathology with stable moderate behavioural abnormalities (Maxan et al., 2020).

2.4 Amyotrophic Lateral Sclerosis (ALS)

19ALS involves the progressive loss of motor neurons, leading to paralysis and eventual respiratory failure. While rodent models have contributed to understanding ALS mechanisms, they lack the complex motor and cognitive features seen in humans. Among the four NDs discussed in this review, the ALS field lacks NHP contributions. It is only since the 43-kDa transactive response DNA-binding protein (TDP-43) was identified as the major component of the protein aggregates and of the insoluble fraction in the brains of patients with sporadic ALS and frontotemporal lobar degeneration (FTLD) that NHP modelling efforts emerged. So far, only AAV-mediated delivery of TDP-43 in various places (some related but some unrelated to ALS or FTLD pathology) has been tested as a putative pathocopy model. One notable model involves the overexpression of human wild-type TDP-43 in the spinal cords of macaques (Uchida et al., 2012). This model successfully replicates key pathological features of ALS, including cytoplasmic mislocalisation of TDP-43, motor neuron degeneration, and muscle atrophy. Importantly, these pathological changes are more pronounced in the lateral nuclear group of Rexed lamina IX, mirroring the selective vulnerability observed in human ALS. Although strangely injected into the substantia nigra, Yin et al. suggested that TDP-43-mediated neuronal pathology likely depends on species-specific factors. They showed that the primate-specific caspase-4 enzyme cleaves mutant TDP-43, leading to its accumulation in the cytoplasm (Yin et al., 2019). Further efforts are required to exploit this primate specificity and allow replication of earlier findings before claiming these preliminary data constitute a primate model of ALS.

3 Advantages and Challenges of NHP Models

20NHPs offer substantial advantages for modelling NDs, particularly due to their close anatomical, genetic, motoric and cognitive similarities to humans, succinctly reviewed above. With regards to proteinopathies, NHPs exhibit more accurate modelling of pathological hallmarks, such as protein aggregation (e.g. Aβ, tau, α-synuclein, TDP-43), synaptic loss, and neuroinflammation. These features are often inconsistently replicated in rodents but appear more faithfully in NHP models due to shared aspects of neuroanatomy, immune function, and brain ageing. Three striking examples among many: First, the primate specificity of certain protein truncation promoting aggregation, i.e. the caspase-4 enzyme cleavage of TDP-43 (Yin et al., 2019). Second: PD is characterised by the degeneration of melanised dopaminergic neurons in the substantia nigra (Herculano-Houzel, 2009) and the presence of Lewy pathology across the entire nervous system. Since monkeys exhibit neuromelanin in substantia nigra dopaminergic neurons, among others traits (Herculano-Houzel, 2009; Teil et al., 2021), they are ecological physiological models (see the work of M. Vila and J. Lanciego around AAV-tyrosinase expression in rodents and macaques). The high anatomical and functional correspondence between NHPs and humans yields information not obtained from rodent models Third: the primate-specific susceptibility to degeneration in response to polymorphs otherwise innocuous in rodents. Injection of PD patient-derived α-syn in NHPs has illustrated the exquisite vulnerability of primate dopaminergic neurons (Recasens et al., 2014). We recently showed that two distinct α-syn pathogenic structures, large or small, have differential effects in mouse and NHP. While large aggregates are toxic in both species, small aggregates induced a comparable extent of nigrostriatal degeneration at the level of nigral dopaminergic neuron cell bodies and striatal dopaminergic terminals in primates (but did not in mice; Bourdenx et al., 2020). More broadly, such results emphasise the importance of NHPs in therapeutic development to not miss the toxicity of certain polymorphs that would have been ignored while relying only on rodents.

21The capacity for longitudinal studies is another key strength of NHP models. Their relatively long lifespans make it possible to track the onset and progression of disease phenotypes over time, mirroring the slow evolution of human neurodegenerative diseases. This temporal resolution allows researchers to observe early pathological events and age-related changes often missed in short-lived animal models. The example of amyloid allometric versus tau isometric pathology progression in NHP compared to humans highlights the complexity of the question.

22Despite their advantages, NHP models face several significant challenges. Ethical and regulatory concerns are at the forefront. The cognitive and social complexity of primates raises moral issues, leading to strict oversight and limitations on their use and adoption of guidelines such as the 3Rs (Replacement, Reduction, and Refinement). The term 'reduction', however, does not simply refer to reducing the number of animals used. It means reducing the number while still ensuring that the hypothesis can be tested adequately, as underpowering a study is a poor use of animals, which leads to inconclusive results. It is also important to use unbiased, blinded evaluation methods and adhere to ARRIVE guidelines (Animal Research: Reporting of In Vivo Experiments; Percie du Sert et al., 2020) for reporting NHP research. Another major limitation is the high cost of housing and maintaining NHP colonies. These animals require specialised care, enriched environments, and extensive veterinary support, resulting in prohibitive financial burdens for many institutions. Moreover, genetically modified NHPs remain limited in availability. While tools like CRISPR/Cas9 have recently improved gene editing in primates, creating stable transgenic lines is technically complex, time-consuming, and raises further ethical issues. Finally, the long lifespan of NHPs, while advantageous for longitudinal research, also means that studies often span several years. This extended timeline can hinder rapid hypothesis testing and therapeutic screening, especially for diseases that take decades to develop in humans.

4 Conclusion

23In summary, while NHPs provide unparalleled opportunities for advancing our understanding of NDs, they require a careful balance of scientific benefit, ethical responsibility, and resource investment. Their high translational value complements and surpasses rodent models, offering insights into disease pathology and treatment responses that closely mimic human conditions. Despite ethical and logistical challenges, ongoing advancements in gene editing, imaging, and alternative research approaches continue to enhance the utility of NHPs in neurodegenerative research. As ethical considerations remain paramount, integrating these models with computational and in vitro methods will ensure responsible and effective research in the future.

Conflict of Interest

24EB is a director and shareholder of Motac Neuroscience Ltd, and the co-founder and shareholder of Treefrog Therapeutics and SE Therapeutics.