Translating Basic Science Into Therapies For PD
Rethinking Parkinson’s Disease Models:
For more than four decades, animal models have been central to Parkinson’s disease (PD) research. They have helped identify molecular pathways, characterize basal ganglia circuitry, and evaluate potential therapies. Yet, despite hundreds of successful preclinical studies, no disease-modifying therapy for PD has emerged from them.
This translational failure does not mean animal models are unimportant. Rather, it forces us to ask whether these models are appropriate and whether they faithfully reproduce the biological processes that drive human PD. The answer appears to be that traditional models excel at studying some aspects of the disease but fall short in capturing those that are critical to the development of a disease-modifying therapy.
The Successes and Failures of Toxin Models
Toxin-based models have dominated the field for decades. Compounds such as 6-hydroxydopamine (6-OHDA), MPTP, rotenone, and paraquat have been used in animal models to kill dopaminergic neurons or their terminals, producing behavioral deficits that resemble parkinsonism. These models have been invaluable for establishing basic principles of nigrostriatal function and for testing symptomatic therapies. Levodopa, dopamine agonists, and deep brain stimulation strategies all benefited from insights obtained in toxin models. However, toxin models have major limitations. First, they typically induce degeneration over days or weeks and fail to mimic the slowly evolving process seen in patients. Second, they largely bypass the pathogenic mechanisms believed to underlie idiopathic PD, including mitochondrial dysfunction, α-synuclein pathology, aging-related changes, and distributed circuit adaptations. Third, toxin administration often results in acute inflammation, which is unlikely to be a disease driver except in rare situations. Fourth, toxin dosing is often not adequately considered as an experimental variable and is critical to the interpretation of results, particularly negative ones.
These shortcomings are not trivial. Every therapeutic strategy that has failed in human clinical trials has been ‘validated’ in a toxin-based model. Treatments targeting oxidative stress, inflammation, excitotoxicity, mitochondrial dysfunction, or trophic signaling have produced encouraging preclinical results only to fail in human clinical trials. To be sure, there are potential explanations for this translational debacle that are un-related to the model per se. Achieving adequate target engagement and biological efficacy is often easier in an animal model than in a human. Nevertheless, this disappointing record has led many of us to question whether acute toxin lesions are capable of predicting therapeutic efficacy in a chronic human neurodegenerative disease. Again, it is not that toxin models are useless. They remain excellent tools for evaluating symptomatic interventions. But protection.
Genetic Models: Mechanistic Insight Without Parkinsonism
The discovery of PD-linked genes generated enormous optimism that genetically engineered mice would provide more faithful disease models. Mutations in SNCA, LRRK2, Parkin, PINK1, DJ-1, VPS35, and other genes were incorporated into mouse genomes with the expectation that they would reproduce the human disorder. These models have unquestionably advanced understanding of PD biology. They have revealed important roles for mitochondrial quality control, lysosomal function, protein aggregation, vesicular trafficking, and cellular stress responses. They have also clarified how mutations perturb neuronal physiology long before cell death occurs. Yet, most genetic mouse models do not develop robust, progressive parkinsonism. Many exhibit subtle physiological abnormalities, biochemical alterations, or modest behavioral phenotypes, but they fail to show the extensive nigrostriatal degeneration and levodopa-responsive motor deficits characteristic of PD.
Why these models have failed is not completely clear. One possibility is that species matters. PD is uniquely human. However, it has been difficult to pinpoint precisely the ways in which human dopaminergic neurons differ from rodent (or primate) dopaminergic neurons that would explain this shortcoming. Another possibility is that time matters. Human PD evolves over decades normally, and it is not too surprising that trying to reproduce it in the short lifespan of a rodent has failed. As a consequence, it is fair to conclude that introducing a disease-causing mutation into a mouse may reproduce some molecular mechanisms while failing to recreate the full syndrome.
The Need for Progressive Models
With a few exceptions, human PD is not the result of an acute injury. It is a decades-long disorder involving progressive changes in neuronal function, connectivity, and network organization. Circuit adaptations unfold gradually and often precede overt neuronal death. Therefore, understanding how motor symptoms emerge requires models that reproduce this temporal evolution. This is where progressive models become particularly valuable. Instead of producing ‘over-nite’ dopamine depletion or mild molecular abnormalities, these models allow us to follow the emergence of pathology from presymptomatic stages through clinically relevant motor dysfunction. They create opportunities to identify the network changes that transform a compensated system into a symptomatic one. Although there are a variety of models that progress over the course of several months from nominally prodromal or non-parkinsonian stages to parkinsonism, among the most compelling examples is the MCI-Park mouse, created by disrupting mitochondrial complex I function selectively in dopaminergic neurons through deletion of the Ndufs2 gene. Importantly, mitochondrial complex I dysfunction is a well-established feature of human PD, giving the model strong biological relevance. Unlike most traditional models, MCI-Park mice develop progressive, levodopa-responsive parkinsonism that includes non-motor deficits. Pathology begins in nigrostriatal axons and terminals before spreading to cell bodies, closely mirroring the “dying-back” pattern observed in human disease. The model also contains an extended prodromal period during which compensatory adaptations can be studied in detail.
Challenging a Core Dogma of Parkinson’s Disease
Perhaps the most important insight gained from study of the MCI-Park model is conceptual. For three decades, the dominant view has been that striatal dopamine depletion is both necessary and sufficient for the emergence of PD motor symptoms. This idea was supported by toxin models, in which severe striatal dopamine loss is tightly linked to parkinsonian motor deficits. But, as pointed out above, the toxin models are not progressive in a sense that is relevant to PD and deplete dopamine throughout the basal ganglia in one fell swoop.
The MCI-Park model challenges this ‘classical’ perspective. In these mice, there is a profound dopamine depletion in the striatum well before parkinsonian motor deficits emerge. In other words, substantial striatal dopamine loss can exist without overt parkinsonism. Symptoms appear only later, after dopamine depletion spread to additional basal ganglia territories. The implication is profound: striatal dopamine depletion may be necessary, but it is not sufficient. Dopaminergic signaling in other basal ganglia structures also appears to play an essential role in maintaining normal movement. Indeed, evidence from this work suggests that restoration of extra-striatal dopamine signaling can ameliorate motor deficits even after parkinsonism emerges. This perspective shifts the focus from a single circuit node to a distributed basal ganglia network. Rather than viewing PD as simply a disorder of striatal dopamine deficiency, it suggests that motor disability arises when compensatory mechanisms throughout the network can no longer offset the progressive loss of dopaminergic modulation across multiple interconnected regions.
One might ask whether the MCI-Park mouse accurately reproduces the sequence of events that lead to clinical parkinsonism in humans. For example, it might not be the case that the profound axon-first staging in this model is truly ‘human-like’. If the goal is to understand PD pathogenesis, this matters. But from the standpoint of understanding the circuit mechanisms underlying symptoms, it doesn’t. It is a tool to help us test hypotheses and has provided the first clear test of the classical network model – which it falsified.
Looking Forward
Animal models are an indispensable tool in PD research. But their greatest value may lie not in predicting whether a therapy will prove disease modifying in the clinic, but in revealing disease mechanisms and identifying the network-level processes that drive the emergence of symptoms. Toxin models have taught us about the importance of dopaminergic signaling in PD. Genetic models have illuminated molecular pathways. Progressive models such as MCI-Park offer us the opportunity to understand how a vulnerable neural system gradually transitions from compensation to disability. If the goal is to develop truly disease-modifying therapies, understanding that transition may be every bit as important as understanding why dopamine neurons die in the first place.
D. James Surmeier, PhD works at Northwestern University as the Nathan Smith Davis Professor and Chair in the Department of Neuroscience at Feinberg School of Medicine. He has spoken at past World Parkinson Congresses. He will be discussing this topic in our September 8, 2026 Research Spotlight. Register for the spotlight HERE and come with your questions!
Ideas and opinions expressed in this post reflect that of the authors solely. They do not necessarily reflect the opinions or positions of the World Parkinson Coalition®