The Future of Parkinson’s Disease Research: A New Paradigm of Human-Specific Investigation Is Necessary… and Possible

Parkinson’s disease (PD) is a complex neurodegenerative condition with a multifactorial origin. To date, approaches to drug discovery for PD have resulted in symptomatic therapies for the motor manifestations and signs associated with neurodegeneration but have failed to identify preventive or curative therapies. This failure mainly originates from the persistence of major gaps in our understanding of the specific molecular basis of PD initiation and progression. New approach methodologies (NAMs) hold the potential to advance PD research while facilitating a move away from animal-based research. We report a workshop involving NAM experts in the field of PD and neurodegenerative diseases, who discussed and identified a scientific strategy for successful, human-specific PD research. We outline some of the most important human-specific NAMs, along with their main potentials and limitations, and suggest possible ways to overcome the latter. Key recommendations to advance PD research include integrating NAMs while accounting for multiple levels of complexity, from molecular to population level; learning from recent advances in Alzheimer’s disease research; increasing the sharing of data; promoting innovative pilot studies on disease pathogenesis; and accessing philanthropic funding to enable studies using novel approaches. Collaborative efforts between different stakeholders including researchers, clinicians and funding agencies are urgently needed to create a scientific roadmap and support a paradigm change towards effective, human-specific research for neurodegenerative diseases without animals, as is already happening in the field of toxicology.

Generating high-quality iPSCs is expensive and time-consuming.
Cost will likely decrease over time PD-iPSC lines commercially available Generating more characterized and sophisticated models of brain organoids Organoids represent an early stage of embryonic development while PD is a late-onset disease.
Inducing an overexpression of aging-related genes (such as progerin) to recapitulate aging and late-onset diseases Direct conversion of aging donors' fibroblasts into neurons to help retain agingrelated transcriptional signatures Cerebral organoids cannot recapitulate disease progression.
Combine with patients' in vivo and postmortem studies Lack of adequate oxygen and nutrient supply and the accumulation of metabolic waste Adding vasculature Combining with fluidic technologies Random and uncontrolled nature of organoid growth Better characterization of models and standardization of culture protocols Organ-chips, MPS (e.g., brain-chip, substantia nigra brain-chip, neurovascular unit-on-a-chip, gut-brainliver-on-a-chip, etc.) To investigate interaction between organs or organ systems relevant to PD, including brain, BBB, intestine, and the gut-brain axis To model the effects of exposure to abnormal αSyn aggregates Performing more physiologically relevant in vitro tests for novel therapies; looking at the influence of the gut microbiota iPSC-based OoCs: studying genetic, ethnic, sex differences, and geneenvironment interactions Applications for personalized medicine Real time monitoring of electrical activity, and physicochemical parameters Lack of robust and reproducible iPSC differentiation protocols for derivation of several cell types of the CNS Limited lifespan of certain types of cells, e.g., endothelial cells

Extensive characterization of OoCs and multi-OoCs
Omics/ Multi-omics approaches (e.g., genomics, transcriptomics, proteomics, metabolomics, exposomics) To define the molecular mechanisms underlying PD pathogenesis Finding biomarkers for clinical detection, progression and treatment Defining genetic and Integration of different omics datasets is a challenging task that relies heavily on data-mining and machine learning algorithms.
Opportunity to establish committed consortia with a multi-disciplinary approach (e.g., combining molecular biology and bioinformatics expertise (Krassowski et al., 2020).

Human cell lines
A better understanding of the molecular and cellular mechanisms underlying PD pathogenesis and progression is essential to developing an effective disease-modifying therapy. Since primary human neural cells, including dopaminergic neurons, are difficult to obtain and maintain in culture, in vitro PD research has been mostly carried-out with permanently established cell lines to date (Falkenburger et al., 2016). Although cell lines obtained from cancerous or in vitro transformed cells, as well as primary neurons isolated from pre-or post-natal rodents, have been useful in the in vitro study of basic mechanisms of PD, they are unlikely to recapitulate all the properties of human neurons in vivo (Napoli and Obeid, 2016;Rich et al., 2020;Harischandra et al., 2020).
Thus, cells with high physiological relevance, good batch-to-batch consistency, and large potential culture scale can provide a human-focused, effective, high-throughput screening model for PD. Conditionally immortalized cells, such as the Lund human mesencephalic (LUHMES) cell line, may offer such possibilities (Fig. 1). LUHMES cells are human embryonic neuronal precursor cells that can be maintained as proliferating cells and can be differentiated into morphologically and biochemically mature dopaminergic-like neurons (Schildknecht et al., 2009;Scholz et al., 2011). They are increasingly used to study the biology of dopaminergic neurons and PD in vitro (Schildknecht et al., 2009;Harischandra et al., 2020;Schildknecht et al., 2015;Loser et al., 2021;Schildknecht et al., 2013b;Gerding et al., 2019). Differentiated LUHMES cells show spontaneous electrical activities and are able to produce/uptake dopamine like human dopaminergic neurons (Zhang et al., 2014).
Coculture of LUHMES cells with astrocytes has elucidated biochemical pathways for the activation of the pro-toxicant MPTP, a pesticide that causes PD-like symptoms by inducing degeneration of nigrostriatal dopaminergic neurons, to toxicant 1-methyl-4-phenyl-pyridinium (MPP+) (Schildknecht et al., 2017;Schildknecht et al., 2015). Known neuroprotective compounds led to different pharmacological profiles in monocultures and co-cultures, suggesting that the presence of glial cells (astrocytes and/or microglia) and the local production of the toxic metabolite MPP+ within the layered cultures play an important role in neuroprotection (Efremova et al., 2015).
LUHMES cell models can be used for investigating αSyn and its role in the development of PD (Paiva et al., 2017;Schildknecht et al., 2013a;Schildknecht et al., 2009). They have also been used to study the effects of oxidative modifications (nitration) on the aggregation of αSyn in vitro as well as its binding to biological membranes. It has been shown that nitrated αSyn stimulates the aggregation of unmodified αSyn molecules by acting as a catalyst for the process (Schildknecht et al., 2013a).
Adverse outcome pathways (AOPs) collate and organize existing information connecting the modulation of a molecular target (molecular initiating event) through a sequence of crucial biological key events to an adverse outcome. LUHMES models have been used to build an (AOP), showing mechanistic plausibility for epidemiological evidence on the relationship between MPTP exposure via binding to mitochondrial complex I with the onset of parkinsonian motor deficits (Terron et al., 2018). The LUHMES model system is a closer approximation of human nigrostriatal neurons than traditional cultures and allows the investigation of the complex biochemical pathways underlying PD development as well as the screening of candidate neuroprotective drugs, promising increased human-relevance and translational success. The ease of culturing these cells allows for the setup of cell-based assays for high-throughput screening. An important challenge to overcome is the continual synthesis and release of dopamine in culture (Schildknecht, unpublished). The cells are difficult to culture for extended periods of time, but this problem can be overcome by culturing them in 3D neuronal spheres, which allows adequate oxygen and nutrient supply for survival of the innermost cells and small molecules penetration. This model could be suitable to identify environmental contributions to PD pathogenesis as well as to address the processes of resilience and recovery in neurotoxicology and PD in future studies (Smirnova et al., 2016).

Brain organoids
A brain organoid is an artificially grown, in vitro miniature organ resembling parts of the brain or defined brain regions (e.g., midbrain organoids) (Fig. 2). Brain organoids are generated by culturing induced pluripotent stem cells (iPSCs), embryonic stem cells, progenitor cells, or multiple types of differentiated cells using specific conditions and media. Organoid brain models allow the simulation of specific changes in neurological disorders and offer several benefits over traditional two-dimensional cultures and whole animal-based approaches in studying PD. The 3D structure increases model complexity over monolayer cultures by allowing more physiologically relevant cell-to-cell interactions and allows cultures to be maintained for longer periods of time. These models are electrophysiologically active, support investigations of increasing complexity, and may be composed of the major cell types in the brain, including neurons, astrocytes, and oligodendrocytes (Smits and Schwamborn, 2020). Human iPSC-derived midbrain organoids are able to produce neuromelanin inclusions that resemble those isolated from the human substantia nigra, in contrast to in vivo rodent models, human midbrain dopaminergic neurons generated using traditional 2D cell culture methods, or midbrain-like organoids generated from mouse embryonic stem cells, which typically do not produce neuromelanin (Jo et al., 2016). Midbrain organoid models are being employed for understanding mechanisms of PD, to decipher risk factors, as well as to identify small molecule drugs to rescue PD pathology (Zanetti et al., 2021;Smits and Schwamborn, 2020).
In addition to their derivation from the species of interest, iPSC-based organoid models can be developed using cells from patients with PD. This strategy enables investigation into genetic differences, sex differences, gene-environment interactions, personalized medicine, and clinical transplantation therapy (Schwamborn, 2018).
Since in vitro organoid differentiation recapitulates the early stages of brain development, brain organoids may be useful in the study of developmental components of PD and the identification of early biomarkers. If the developmental theory and the "multiple hit" concept are correct, an early identification of the deregulated developmental processes may allow intervention therapy in at-risk individuals before symptoms appear.
Human midbrain-specific organoids derived from PD patients carrying a specific LRRK2 mutation were used to investigate PD-relevant pathogenetic mechanisms (Smits et al., 2019). These organoids recapitulated key characteristics of the disease, including decreased amounts of midbrain dopaminergic neurons compared to isogenic control organoids. This pinpoints an impaired midbrain dopaminergic neuron specification during organoid development, supporting the hypothesis that PD has a neurodevelopmental component (Smits et al., 2019). Further studies support the usability of midbrain organoids for PD in vitro disease modeling (Kwak et al., 2020;Nickels et al., 2020;Mohamed et al., 2021;Kim et al., 2019;Zanetti et al., 2021;Monzel et al., 2020;Jo et al., 2021).

Fig. 2: 30-day old midbrain organoid
The neuronal marker MAP2 is shown in red, the dopaminergic neuron marker TH is shown in green, and the DNA in cell nuclei is shown by Hoechst staining in blue. Image provided by Mudiwa Nathasia Muwanigwa (University of Luxembourg).
iPSCs differentiated from parkin (PRKN)-mutated patients into midbrain organoids confirmed decreased numbers of astrocytes in PRKN-mutated organoids compared with age-and sex-matched controls. This correlates with observations in human brains from individuals with PRKN mutations compared with subjects with idiopathic PD or healthy individuals and suggests the probability of an astrocyte-associated cell death mechanism for dopaminergic neurons in the brains of PRKN-mutated patients (Kano et al., 2020).
Inducing an overexpression of aging-related genes (such as progerin) in iPSC-derived organoids is a technique that has been applied as an artificial aging system to recapitulate aging and late-onset diseases like PD (Hu et al., 2018;Miller et al., 2013). Alternatively, the direct transdifferentiation of aging donors' fibroblasts into neurons (namely induced neurons, or iNs), circumventing cell reprogramming toward an embryonic phenotype, has shown promise: iNs were found to retain aging-related transcriptional signatures when compared to iPSCs and their neuronal derivatives (Mertens et al., 2015). Miniature brain organoids also allow the characterization of cellular responses through electrophysiology, immunocytochemistry, histology, flow cytometry, and gene expression, and to associate phenotypic changes to a drug's mechanism of action more rapidly and accurately than ever before. Certain types of brain organoids allow human-relevant oligodendrocyte myelination. Examples of commercially available 3D miniature brain organoids are BrainSim® 1 by AxoSim, which are composed of neurons, astrocytes, and oligodendrocytes, and human midbrain-organoids by OrganoTherapeutics 2 . Functional activity of neural organoids can be measured using commercially available MEA systems such as by Axion Biosystems 3 .
Limitations of organoid models include the lack of a vasculature and the random and uncontrolled nature of their growth. Future studies will involve generating better characterized and more sophisticated models of brain organoids, including diverse brain regions with immune cells and vasculature, as well as their combination with additional organ-specific organoids such as the gut, which may also allow the study of the impact of gut microbiota on disease development (Reiner et al., 2021). These models will allow the study more extended patient cohorts and the evaluation of potential therapeutics.

Organ-chips
Οrgan-chips are multi-channel 3D microfluidic cell culture devices that model the physiology, tissue-tissue interfaces, and microenvironment of functional units within human organs, as well as inter-organ communication (Bhatia and Ingber, 2014). Organchips are made of transparent flexible polymer, the size of a memory-stick, typically containing two fluidic microchannels separated by a porous flexible membrane that can be coated with extracellular matrix proteins and seeded with cells on either side. Media flow can be precisely controlled, providing nutrients as well as the ability to collect the effluent for analysis (e.g., proteomics or metabolomics). The fluidic channels may be positioned between two vacuum chambers, which can be used to control cyclic stretch in organ-chips such as a gut-on-a-chip to recreate the mechanical forces cells experience inside the body during peristalsis (Kasendra et al., 2018). Organ-chips can be used to investigate interactions between organs or organ systems relevant to PD, including brain, bloodbrain-barrier (BBB), intestine, and the gut-brain axis. Pediaditakis et al. (2021) used Emulate organ-on-a-chip technology to develop a human brain-chip representative of the substantia nigra region of the midbrain comprising iPSC-derived dopaminergic neurons, immortalized microglia, primary astrocytes, primary pericytes, and iPSC-derived microvascular brain endothelial cells, recreating a neurovascular unit (Fig. 3). To model the effects of exposure to abnormal αSyn aggregates and assess the capability of the substantia-nigra brain-chip to generate clinically relevant 1 https://axosim.com/brainsim/ 2 http://organo-therapeutics.com/ 3 https://www.axionbiosystems.com/applications/neural-activity/neural-organoids  endpoints, a model of synucleinopathy was built by adding exogenous human αSyn fibrils into the brain channel. This model was able to replicate many crucial aspects of PD, including accumulation of phosphorylated αSyn, uptake of αSyn fibrils by neurons, astrocytes and microglial cells, mitochondrial dysfunction, reactive oxygen species (ROS) production, neuroinflammation, neuronal cells death, and impaired BBB function. Interestingly, transcriptomic analysis of the endothelial cells in the brain-chip revealed the upregulation of genes involved in inflammation, oxidative stress, autophagy, efflux system, and extracellular matrix deposition and the downregulation of genes that encode for tight junction proteins, suggesting a new pathogenic mechanism and a potential target for therapeutic intervention in PD. Trehalose-treated brain-chips showed reduced neuroinflammation and BBB permeability, rescuing the damaged tight junctions following exposure to αSyn fibrils (Pediaditakis et al., 2021). In summary, the brain-chip provides a promising model for the study of the specific disease mechanisms, including the dynamics of BBB dysfunction and may be useful to characterize the response to PD therapies and identify and evaluate associated biomarkers of the disease.
Nerve-on-a-chip platforms of iPSC-or primary culture-derived 3D organoids allow nerve conduction analyses to measure changes in electrophysiological properties in response to stimuli. Examples of commercially available organs-on-a-chip relevant to PD research are the NerveSim® platform 4 by Axosim and BBB/Brain-on-a-chip by EmulateBio 5 and by ElveFlow 6 (Elveflow 2022).
With regard to genomics, the most recent large-scale genome-wide association studies (GWAS) have discovered more than 90 common variants for sporadic PD risk, age at onset and progression across more than 80 genomic regions Foo et al., 2020;Iwaki et al., 2019;Nalls et al., 2019). Recently, a PD GWAS browser tool 7 was created to support the PD research community with the prioritization of genes for follow-up functional studies to detect potential therapeutic targets (Grenn et al., 2020).
Transcriptomics has the potential to provide substantial insight into disease processes. Transcriptome-wide association studies (TWAS) in bulk human brain tissues have found numerous molecular signatures associated with the disease (Borrageiro et al., 2018). Although these studies have the potential to shed light on PD pathogenesis, they are limited by heterogeneous cell type composition of bulk tissue samples, and RNA post-mortem degradation (Nido et al., 2020). Performing transcriptomics studies on human-relevant in vitro models may help overcome some of these issues. For example, a single-cell transcriptomics study on human in vitro models allowed the identification of numerous neuronal subtypes with transcriptionally different profiles and sensitivities to stress, showing cell type-specific perturbations in human dopaminergic neurons. These findings have implications for cell replacement therapies (Fernandes et al., 2020).
The role of protein modifications during PD pathogenesis could be better understood by the extensive application of proteomics approaches. Several proteomic analyses performed on brain tissue, cerebrospinal fluid, and blood of PD patients have detected a wide spectrum of protein modifications underlying disease pathogenesis (Dixit et al., 2019;Posavi et al., 2019;Rotunno et al., 2020;Werner et al., 2008). These studies are limited by the types of brain areas or biofluids accessible for research.
A growing amount of research has demonstrated that metabolomics profiling holds great promise in providing insights into molecular pathogenesis and could be very effective in detecting candidate biomarkers for clinical detection, progression and treatment of PD, as well as in the study of the influences of environmental risk factors (Dong et al., 2018;Goldman et al., 2018;Peters et al., 2019;Sinclair et al., 2021;Troisi et al., 2019).
Omics approaches can also be applied to study gut microbiota composition and functions in PD patients. Gut microbiome alterations in PD have been reported frequently (Mao et al., 2021;Romano et al., 2021). A recent study showed that fecal microbiome and metabolome composition in PD was considerably different from controls, supporting a role for microbial metabolites as potential targets for developing new biomarkers and treatments in PD (Tan et al., 2021).
Specific PET tracers allow distinction of PD from overlapping diseases, such as multiple system atrophy, and can also be used as a biomarker for dopaminergic neuron degeneration before symptom onset. This could allow the testing of therapies to prevent or halt PD before irreversible symptoms present (Loane and Politis, 2011). The ability to gather quantitative longitudinal data relative to biomarkers makes non-invasive imaging techniques essential to current and future clinical trials (Saeed et al., 2017). Development of a PET tracer for αSyn is underway (Korat et al., 2021) and promises to be a game changer like it was for amyloid and tau PET tracers in AD. In this context, the recent NIH initiative "Center Without Walls for Imaging Proteinopathies with PET" engages several academic and industrial centers with different areas of expertise to develop imaging biomarkers.
As for many NAMs, new imaging technologies will be required to overcome excessive operational costs. Moreover, investigators will need to increase the precision of the identifying correlations between measures and PD-related clinical manifestations, as well as formulate procedures to deal with complex data sets to improve quantification of imaging features. Despite current limitations, existing imaging technologies represent important tools to assist human-relevant PD research approaches and should be employed on a wide scale in clinical studies to improve correlations between imaging measurements and clinical manifestations.
Further progress in non-invasive human brain imaging could also reduce the reliance on conventional pathological tissue analysis in inadequate animal disease models.

Computational models
A major and often underestimated challenge is the translation of preclinical neuroprotective findings of putative therapeutics to the human patient. The ideal drug must achieve sufficient target exposure at the site of action, with limited off-target effects, a long enough duration to actively engage the target, and needs to be devoid of toxicity. For compliance reasons, the formulation should ideally be orally once-a-day, although PD patients are generally open to other, more invasive approaches. In addition, any new drug needs to have minimal pharmacokinetic (PK-PK) or pharmacodynamic (PD-PD) interactions with standard-of-care medications in PD patients.
Physiologically based pharmacokinetic (PBPK) and pharmacodynamic (PDPK) modeling is suitable to define the kinetics and dynamics of compound exposure at the site of action and to simulate the interaction of pharmacological agents with transporters at the BBB to derive an estimate of brain or cerebrospinal fluid penetration.
Quantitative systems pharmacology (QSP) is a relatively new approach aiming to predict pharmacodynamic effects and to assess the therapeutic potential of novel compounds, taking individual patient characteristics into account. For PD, this approach is based on mechanistic and biology-informed modeling of the interaction between a compound and neuronal circuits to simulate the pharmacodynamic effects of a drug in conjunction with patient-specific genotypes, concomitant medications, and disease conditions on functional clinical scales (Geerts et al., 2013(Geerts et al., , 2020. To discover novel targets for PD treatment and potential synergistic pharmacodynamic effects between different therapeutic compounds, Roberts et al. (2016) built a computer-based QSP platform of the closed cortico-striatal-thalamic-cortical basal ganglia loop of the dorsal motor circuit. Results indicate that the platform can achieve a strong correlation with past clinical trials, including therapeutic interventions that did not show clinical benefit (Roberts et al., 2016). This approach allows an estimation of the impact of individual standard-of-care medications, not only on possible biomarkers but also on a clinically relevant functional outcome such as Unified Parkinson's Disease Rating Scale (UPDRS) or off-time. In addition, the platform is patient-centric as it is calibrated using historical clinical trials. Taken together, PBPK and QSP modeling allow for optimal clinical trial design to increase the probability that therapeutic interventions with potential in preclinical assays demonstrate a clinically relevant signal in clinical trials. The main limitation of computational models is that results are based only on existing knowledge and input data, and, consequently, feeding and updating models using novel, emerging, human-relevant data is crucial.

Future directions -How can we improve PD research?
Many years of high-cost failures (ethical and financial) in translating research findings and treatment candidates from preclinical (animal) disease models to human benefits/ therapeutic use demand a serious reconsideration of the confidence placed in animal models in PD research. In addition to the failure in drug development associated with the use of inadequate preclinical (both in vivo and in vitro) models, which have questionable relevance to humans, the use of these models in an excessively reductionist approach to dissect the possible contributions of single gene(s), protein(s) or mechanism(s) to the onset of a complex, multi-factorial human condition such as PD might have contributed to the translational failure. For these reasons, some key issues were highlighted by the workshop participants that should be seriously considered, with the aim to refocus current and future research strategies and priorities in PD research.

Integration of new approach methodologies to enhance Parkinson's disease research
Human-specific NAMs should be used to investigate, establish, and explain disease processes at multiple levels of biological complexity, from the molecular level to cells/tissues/organs, to organism/population levels (Fig. 4). Combining novel human-based cellular and computational models with non-invasive imaging techniques, epidemiological and clinical data would enable humanrelevant data discovery, while moving away from animal-based PD research. This approach would provide multidimensional insight into the association between PD pathogenetic processes and the influence of the environment on them, allowing the acquisition of comprehensive and holistic biological information in a human context. Merging different organoids to generate complex multiorgan-on-a-chip (multi-OoC) systems has the potential to transform medical research by opening new avenues for understanding multi-organ diseases and for developing personalized treatments (Picollet-D'hahan et al., 2021). For example, a gut-brain axis-on-a-chip (Kim et al., 2021) and a multi-organ-on-a-chip system including brain, liver, and intestine (Trapecar et al., 2021)  monitoring of tissue constructs (Ferrari et al., 2020;Liang and Yoon, 2021) holds promise for the advancement of neurodegenerative disease research.
Genetic factors, gene expression, proteins linked with disease or signaling pathways, post-translational modifications, relevant metabolites, and the microbiome can be studied with a wide array of omics approaches, next-generation sequencing technologies, gene expression profiling, mass spectrometry, and magnetic resonance spectroscopy (Fig. 4). These, together with integrated computational modeling, are already laying the groundwork for a systems-biological understanding of disease etiopathology without the use of animals, and thus without the problem of interspecies differences. Moreover, multi-omics approaches may enable the stratification of patients into meaningful disease subgroups, resulting in improved disease management and the development of effective treatments (La Cognata et al., 2021).
Human ex vivo materials derived from patients and healthy subjects, such as CSF, plasma and blood samples, post-surgical biopsies, and post-mortem tissues can address population/body biological complexity and be valuable to find early biomarkers. Collecting comprehensive data from both healthy individuals and PD patients is critically important to validate/inform human-based in vitro and in silico models and to advance PD research. The Michael J. Fox Foundation 8 (MJFF) (Marek Kenneth, 2011) collects clinical, imaging, and genetic data from PD patients and healthy people to provide the research community with critical human-relevant insight into the lived experience, genetics, and variability of PD. Similarly, the ADNI (Alzheimer's Disease Neuroimaging Initiative), has fostered tremendous progress regarding plasma biomarkers and longitudinal personalized trajectory of brain pathology and clinical outcomes of AD over the last 20 years (Veitch et al., 2021).
The application of an analytic approach like the adverse outcome pathways (AOPs) framework (Paini et al., 2021) could help to better categorize existing knowledge about the factors and events associated with PD, to make gaps and overlaps in the knowledge of the disease more evident, and to suggest novel research strategies. The AOP 9 framework allows the integration of many types of information at different levels of biological organization, from molecular to population scale, providing coherent, biologically based theories and hypotheses to predict the outcome of initiating events. The construction of PD related-AOPs has begun (Terron et al., 2018). Intensive efforts to integrate existing PD data to build upon existing AOP efforts could play a substantial role in transforming our ability to identify effective preventative therapies (Marshall and Willett, 2018b). Further PD-associated pathways exist in the Kyoto Encyclopedia of genes and genome database 10 .
An outline of the available NAMs, novel tools and readouts available to aid the design of human-oriented PD research, accounting for multiple levels of complexity, is depicted in Figure 4.

Lessons from Alzheimer's disease research
There was consensus among the experts that we should learn from recent approaches in AD research to inform a PD strategy. More than 400 human trials of potential AD treatments have been conducted but almost no new drugs have reached the market (King, 2018). In AD, one of the major sources of researchers' concern is the poor human relevance of the animal models used in the preclinical stages of drug development and in biomedical research (Pistollato et al., 2016). Moreover, there is an increasing awareness that a paradigm shift toward human-focused, rather than animal-focused research is required to better clarify AD pathogenesis and to enable the design of more effective treatments and preventive interventions (Cavanaugh et al., 2014;Pistollato et al., 2015Pistollato et al., , 2020 Analogous to the effort in AD research, PD research is now beginning to focus on early modification and prevention of disease, based on CSF or plasma biomarkers that may predict onset of disease before symptoms present (Smedinga et al., 2021;Wu et al., 2011). Identifying patients who are on a trajectory to develop the disease would allow for early interventions that might slow-down or ideally halt the disease process. This strategy shift comes with new challenges that have hindered progress in the field of AD in recent years, where an analogous research strategy led to a number of disappointing trial results. Capitalizing on lessons and experiences from AD research in the field of PD will accelerate progress in PD research (Smedinga et al., 2021). Also, it has become clear that progression in PD may be related to mixed brain pathologies (Buchman et al., 2019).

Data sharing
For science to work effectively and for society to obtain the full benefits of scientific efforts, it is crucial that scientific data are made freely available. As mentioned above and inspired by the ADNI, the MJFF promotes knowledge-sharing and collates scientific findings on individuals (including de novo PD patients, unmedicated PD-, prodromal-, and at-risk patients both with genetic and sporadic PD, and healthy controls). They collaborate with the research community to collect and distribute data resources. The Parkinson's Progression Markers Initiative 11 , sponsored by the MJFF 8 , is an ongoing longitudinal observational study launched in 2010 that collects comprehensive clinical, imaging, and genetic data, and biological samples, with more than 1,500 participants. This study is supported by a number of industry partners and has built a robust database and biorepository, which is made accessible to the research community to promote biomarker identification and validation as well as therapeutic development. Similarly, the Aligning Science Across Parkinson's 12 (ASAP) initiative is fostering collaboration and resources to better understand the underlying causes of PD by supporting open-access policy. More initiatives like these should be promoted.
Another important fact that emerged during the workshop discussion concerned the importance of sharing and publishing negative data from animal studies. These data need to be accessible to avoid experimental replication, to prevent wasting time and resources, and to ensure that the best, most informed, evidence-based research is funded and conducted.

Funding and stakeholder engagement
There is a huge need for innovative, novel approaches and trials, including unbiased "fishing expeditions" to find out more about disease pathogenesis. This type of research approach is essentially data-driven or observational research, and its primary objective is to seek meaningful insights, novel mechanisms, patterns or correlations within the data or subject of observation.
The experts mentioned "risk aversion" to novel concepts as a reason for a lack of robust funding to test novel ideas for PD research. This is where philanthropy can play an important role: Philanthropic funding can encourage innovative research by supporting studies that might not otherwise be funded and by promoting a more collaborative and open approach to research.
A scientifically driven transition towards a human-based paradigm in toxicology is taking place, following the roadmap initially traced by the landmark report "Toxicity Testing in the 21st Century: A Vision and a Strategy" published in 2007 by the U.S. National Research Council (NRC, 2007). A similar paradigm shift appears to be necessary in biomedical research, in particular in the field of neurodegenerative diseases, including PD, where the lack of translatability from animal disease models to human patients is particularly evident and, at the same time, the current available human-based models and their future potential applications are very promising. A collaborative effort between different stakeholders including researchers, clinicians, and funding agencies, is urgently needed to create a scientific roadmap and support a paradigm change towards effective, human-specific PD research without animals.
Key recommendations to improve PD research are summarized in Table 2.
Tab. 2: Key recommendations to improve PD research 1 Combining the use of different NAMs and large-scale epidemiological studies accounting for all levels of complexity, from molecular, to cell/tissue/organ, to organism, to population level, while moving away from animal-based research • Integrating different human-based models and tools (e.g., simpler with more complex models, organoids with fluidic technologies, multi-OoC-platforms, human-on-chip, multi-omics approaches); • Combining NAMs with large scale epidemiological studies to account for all levels of complexity; • Leveraging the AOP and PPMI concept to structure existing knowledge about the factors and events concerned with PD 2 Learning from recent approaches in AD research to inform PD strategy • Identify patients who are on a trajectory to develop disease for early intervention; • Learn from mistakes: decades of animal-based studies in AD did not result in an effective cure; 3 Sharing data • Science data should be openly available. Need for robust databases accessible to the research community to promote biomarker identification/validation and therapies development; • Sharing and publishing negative data on animal studies to avoid experiments' replications making sure that the best, most informed resources are done; 4 Promoting unbiased "fishing expeditions" research to find more about disease pathogenesis; 5 Focusing on philanthropic funding that has the ability to take up the risk and support studies that might not otherwise be funded: Risk aversion to novel concepts often results in a lack of robust funding to test new ideas or PD research. 6 A collaborative effort between different stakeholders including researchers, clinicians, and funding agencies is urgently needed to create a scientific roadmap and support a paradigm change towards effective, human-specific research for neurodegenerative diseases without animals, similar to what is already happening in the field of toxicology.