The STRAT-PARK cohort: A personalized initiative to stratify Parkinson ’ s disease

The STRAT-PARK initiative aims to provide a platform for stratifying Parkinson ’ s disease (PD) into biological subtypes, using a bottom-up, multidisciplinary biomarker-based and data-driven approach. PD is a heterogeneous entity, exhibiting high interindividual clinicopathological variability. This diversity suggests that PD may encompass multiple distinct biological entities, each driven by different molecular mechanisms. Molecular stratification and identification of disease subtypes is therefore a key priority for understanding and treating PD. STRAT-PARK is a multi-center longitudinal cohort aiming to recruit a total of 2000 individuals with PD and neurologically healthy controls from Norway and Canada, for the purpose of identifying molecular disease subtypes. Clinical assessment is performed annually, whereas biosampling, imaging, and digital and neuro-physiological phenotyping occur every second year. The unique feature of STRAT-PARK is the diversity of collected biological material, including muscle biopsies and platelets, tissues particularly useful for mitochondrial biomarker research. Recruitment rate is ~150 participants per year. By March 2023, 252 participants were included, comprising 204 cases and 48 controls. STRAT-PARK is a powerful stratification initiative anticipated to become a global research resource, contributing to personalized care in PD.


A B S T R A C T
The STRAT-PARK initiative aims to provide a platform for stratifying Parkinson's disease (PD) into biological subtypes, using a bottom-up, multidisciplinary biomarker-based and data-driven approach.PD is a heterogeneous entity, exhibiting high interindividual clinicopathological variability.This diversity suggests that PD may encompass multiple distinct biological entities, each driven by different molecular mechanisms.Molecular stratification and identification of disease subtypes is therefore a key priority for understanding and treating PD.STRAT-PARK is a multi-center longitudinal cohort aiming to recruit a total of 2000 individuals with PD and neurologically healthy controls from Norway and Canada, for the purpose of identifying molecular disease subtypes.Clinical assessment is performed annually, whereas biosampling, imaging, and digital and neurophysiological phenotyping occur every second year.The unique feature of STRAT-PARK is the diversity of collected biological material, including muscle biopsies and platelets, tissues particularly useful for mitochondrial biomarker research.Recruitment rate is ~150 participants per year.By March 2023, 252 participants were included, comprising 204 cases and 48 controls.STRAT-PARK is a powerful stratification initiative anticipated to become a global research resource, contributing to personalized care in PD.
Parkinson's disease (PD) is the most rapidly growing neurodegenerative disorder, and a major cause of disability and mortality on a global scale (Bloem et al., 2021).Current treatments for PD are purely symptomatic, and trials of potential disease-modulating therapies have been unsuccessful (Athauda and Foltynie, 2015).
Efforts to understand and treat PD are undermined by its vast clinicopathological heterogeneity.While referred to as a single entity, PD is a heterogeneous syndrome, defined by phenotypical features (Postuma et al., 2015;Postuma and Berg, 2017).Patients exhibit high variability in age of onset, type and severity of clinical features, rate of progression, response to treatment, and underlying neuropathology (Chen-Plotkin et al., 2018;Greenland et al., 2019;Halliday and McCann, 2010).The clinicopathological diversity of PD has led to the hypothesis that distinct biological subtypes of the disease may exist, driven by different molecular mechanisms (Espay et al., 2017b(Espay et al., , 2017a)).Without stratification, such molecular heterogeneity dilutes the biological signal in observational and interventional studies, limiting both mechanistic and therapeutic breakthroughs.Thus, successful molecular stratification of PD and identification of disease subtypes is a key priority, as this will provide insights into individualized progression and prognosis, and most importantly, enable personalized treatment strategies (Berg et al., 2014;Espay et al., 2017a;Rodríguez-Violante et al., 2017).
Current stratification approaches for PD rely mostly on clinical features, with motor phenotype being the most common (Berg et al., 2021;Campbell et al., 2020;Foltynie et al., 2002;van Rooden et al., 2011).For example, the tremor dominant (TD), postural instability and gait disorder (PIGD), or akinetic rigid (AR) motor subtypes have been extensively used (Erro et al., 2020;Jankovic et al., 1990;Qian and Huang, 2019;Thenganatt and Jankovic, 2014).More recently, non-motor symptoms have gradually gained importance in phenotypical subtyping of PD (Fereshtehnejad et al., 2015).Another approach, using multimodal functional neuroimaging, has led to the hypothesis that two forms of PD can be discerned, depending on whether neurodegeneration starts in the brain or the gut (Horsager et al., 2020).Neurotransmitter-centric disease classifications have also been suggested, correlating motor and/or non-motor symptom constellations with dysfunction and degeneration of specific neuronal populations, such as noradrenergic, cholinergic, or serotonergic (Berg et al., 2021).While the above classification schemes support the view that molecular subclasses of PD may exist, no such subtypes have been established to date.In fact, it has been proposed that a bottom-up strategy may be a more effective approach.Rather than starting from clinical classifications and seeking molecular associations for these, this strategy would stratify PD according to distinctive molecular features, and subsequently characterize the clinical phenotype of the emerging subtypes.This approach may prove more fruitful in defining disease subtypes amenable to specific therapies (Espay et al., 2017b).
A molecular stratification approach for PD may initially focus on known molecular processes associated with the disease.While the molecular pathogenesis of PD remains largely unknown, a number of associated biological processes have been identified; mainly aberrant proteostasis, mitochondrial dysfunction, and neuroinflammation (Bloem et al., 2021;Kalia and Lang, 2015).The aggregation of α-synuclein is at the center of most current pathogenic models for PD.Aberrant proteostasis, in the form of lysosomal and proteasomal dysfunction, is believed to contribute to aggregation of α-synuclein in Lewy pathology, a hallmark feature of PD (Dickson, 2012).This is corroborated by the fact that loss of function variants in the GBA gene, encoding the lysosomal enzyme glucocerebrosidase, are associated with increased risk of PD.However, GBA risk variants are only found in 7-15% of PD cases (Ren et al., 2022) and there is no robust evidence for impaired proteostasis pathways in the remaining cases.
Mitochondrial dysfunction, mainly in the form of complex I deficiency and aberrant mitochondrial DNA (mtDNA) maintenance, is considered integral to PD (Chen et al., 2019;Flønes and Tzoulis, 2022).Evidence of these mitochondrial defects has been detected in neurons from multiple brain regions (Flønes et al., 2018) and in peripheral tissues, primarily muscle and/or platelets, and to a lesser extent blood cells (Subrahmanian and LaVoie, 2021).However, the results of previous studies have been variable and, in part, contradictory, raising the pertinent question of whether impaired mitochondrial function is a pervasive feature of PD (Flønes and Tzoulis, 2022;Subrahmanian and LaVoie, 2021).Finally, neuroinflammation involving microglial activation and/or dysfunction has been implicated in PD (Hirsch and Hunot, 2009), but its role remains largely obscure.
Importantly, while the aforementioned processes are associated with PD, it remains unknown whether they are pervasively involved in all cases, or only contribute to specific patient subpopulations.Thus, molecular biomarkers targeting these pathways may help us stratify PD and identify molecular subtypes.Moreover, unbiased molecular datasets, such as multi-omics, may enable the discovery of novel molecular pathways, or even allow hypotheses-free, data-driven stratification to take place.
The main goal of the STRAT-PARK initiative is to establish a longitudinal population-based cohort of sufficient size and with appropriately diverse biosampling and measures, to allow resolving the biological heterogeneity of PD and developing biomarkers for patient stratification in clinical practice.

Study design
STRAT-PARK is a prospective, longitudinal, multicenter cohort study.A total of 2000 individuals, including individuals with PD (n=1,500) and neurologically healthy, demographically matched controls (n=500) will be prospectively recruited and followed up at three clinical centers: Haukeland University Hospital (HUS), Bergen, Norway, St. Olav's University Hospital, Trondheim, Norway, and The London Movement Disorders Centre (LMDC), Ontario, Canada.Together, the three sites have an estimated candidate population of at least 4,000 patients at any given time.The current recruitment rate is 150 participants per year.Controls are recruited among the patients' partners, as well as other volunteers, and must have no signs of neurodegenerative or neuropsychiatric disorders.After initial screening and assessment at baseline, participants are followed prospectively with yearly visits until death, drop-out or study discontinuation.The frequency and content of the visits are summarized in Table 1.The study design is illustrated in Fig. 1.

Inclusion and exclusion criteria
To be eligible for participation, patients must have a diagnosis of clinically established or probable PD according to the Movement Disorders Society (MDS) diagnostic criteria for PD (Postuma et al., 2015), and be able to consent and participate in the study.Participants recruited in Norway must in addition have 123I-Ioflupane single photon emission computed tomography [ 123 I]FP-SPECT (e.g., dopamine active transporter (DAT) scan), or 3,4-dihydroxy-6-18F-fluoro-L-phenylalanine ( 18 F-FDOPA) PET scan confirming nigrostriatal degeneration.Controls are eligible for participation if they are able to consent and participate in the study, and lack signs of parkinsonism, or other neurodegenerative disorder at screening/baseline.A detailed list of participation criteria is given in the study protocol.

Recruitment procedure and centers
STRAT-PARK aims to recruit a disease population which is as unselected as possible and, therefore, representative of the general disease population.Subjects are recruited from the outpatient departments of the study centers "as they present" for first referral or follow-up.In the Norwegian health care system, all individuals suspected to have PD are referred to neurological evaluation and follow up by a public neurology clinic.The Norwegian sites involved in the study are primary referral centers for PD for their respective regions and, therefore, see virtually all newly diagnosed cases in the population.LMDC is a tertiary referral center serving a large population of southwestern Ontario.Patients are referred by family doctors directly, while community neurologists tend to refer patients in need of advanced treatment.Thus, the patient group recruited there is more selected and less representative of the general population.Healthy controls are recruited among the partners or close relatives of the patients and other volunteers.

Procedures
STRAT-PARK participants undergo comprehensive clinical assessment annually with biosampling and imaging every second year.Examinations and biosampling take place in the "ON" dopaminergic state to minimize patient discomfort, except from motor assessment and qEEG at the London site, which is conducted both in the "OFF" and "ON" states.The frequency and content of the visits is summarized in Table 1.

Clinical assessments and questionnaires
Medical history including comorbidities, family history, and current medication is recorded.Environmental and occupational exposures such as smoking and alcohol intake, pesticide exposure and others are documented.Vital signs and body metrics are recorded.
Motor functions are assessed using the International Parkinson and Movement Disorder Society Unified Parkinson's Disease Rating Scale (MDS-UPDRS) (Goetz et al., 2008), part III.At the Norwegian sites, motor assessment is conducted in the "ON" state.At LMDC, motor assessment is performed in both "ON" and "OFF" state.
Non-motor symptoms are assessed using the relevant items in the MDS-UPDRS and specific scales, including the International Parkinson and Movement Disorder Society Non-Motor Rating Scale (MDS-NMS) (Chaudhuri et al., 2020) and the Montreal Cognitive Assessment (MoCA) (Nasreddine et al., 2005).Autonomic dysfunction is assessed with the Scales for Outcomes in Parkinson's Disease -Autonomic Dysfunction (SCOPA-AUT) questionnaire (Visser et al., 2004) and measurement of orthostatic blood pressure (OBP) (Gibbons et al., 2017).Gastrointestinal function is assessed using the Gastrointestinal Dysfunction Scale for Parkinson's Disease (GIDS-PD) (Camacho et al., 2021).Sleep changes are assessed with the REM Sleep Behavior Disorders Screening Questionnaire (RBDSQ) (Stiasny-Kolster et al., 2007) and Parkinson´s Disease Sleeping Scale (PDSS-2) (Trenkwalder et al., 2011).Quality of life is assessed using the 5-level EQ-5D version (EQ-5D-5L) (Herdman et al., 2011).Olfactory function is measured with the Brief Smell Identification Test (B-SIT) (Doty et al., 1996).Administration and rating of non-motor scales and testing for cognition is done in the dopaminergic "ON" state at all sites.

Digital and neurophysiological biomarkers
The LMDC site performs objective motor assessment.They use full body wearable wireless inertial sensing-based motion capture system using Xsens MVN suit to quantify the cardinal motor symptoms of bradykinesia and tremor.The KINARM Endpoint robot, which comprises a robotic manipulandum (robot handle) coupled with a Virtual Reality (VR) display, is used to extract the upper-limb kinematic data which in turn characterizes the PD-related upper limb motor impairments.Finally, gait assessment is done using the Zeno walkway (Zenometrics LLC, Peekskill, NY) in conjunction with the ProtoKinetics Movement Analysis Software (PKMAS).These data are captured and collated via the PKMAS software, resulting in numerous spatial, temporal and pressure-related gait parameters.These parameters are compared with the gait variables derived from the inertial sensing-based motion capture system.The Norwegian sites will implement wearable sensors during 2024, with an emphasis on remote monitoring of physical activity and motor symptoms.
Speech and eye movement are examined at LMDC.Speech is recorded using a head-mounted microphone (AKG-c520, AKG Acoustics, USA) and a digital recording device (Zoom H4nPro, Zoom Corporation, USA) as described (Boutsen et al., 2023).Eye movement examination is done using a screen-based eye tracking device (Tobii Pro fusion, Tobii, Stockholm) as described (De Kloe et al., 2022).During the test, the participant's eye movements, including pursuits and saccades in both horizontal and vertical direction and opto-kinetic nystagmus, are recorded using a visual stimulus.Quantitative cortical (surface) electroencephalography (qEEG) is performed at LMDC in the resting condition with eyes-closed using a standard 32-channels wireless biosignal acquisition system that is recorded with a 256 Hz frequency sampling rate.The qEEG parameters is recorded both in the dopaminergic "ON" and "OFF" state.

Imaging
Imaging is performed every second year throughout the study.Magnetic resonance imaging (MRI) is conducted on a 3 Tesla (3 T) MR-PET scanner at HUS (Biograph mMR, Siemens Healthineers, Erlangen, Germany) and LMDC (3 T Discovery MR750; GE Medical Systems, Milwaukee, Wisconsin using a 32-channel head coil), and a 7 Tesla (7 T) scanner at St. Olavs University Hospital (SIEMENS MAGNETOM Terra scanner).High-resolution 3D T1-and 2D T2-weighted and T2-fluid attenuated inversion recovery (T2-FLAIR) sequences are performed for global and regional cerebral volume measurements.Diffusion tensor imaging (DTI), acquired with 60 directions spread across two b-value shells (1000 and 2500 s/mm 2 ) and opposite phase images for distortion correction, is used for mapping brain water diffusion and assessing changes in cellularity using fractional anisotropy (FA) and structural connectivity (tractography).Neuromelanin imaging using magnetization transfer contrast Orthostatic blood pressure -BP and pulse in the supine position after 5 min rest.
-BP and pulse after 1 min standing.
-BP and pulse after 3 min standing.
All sites X X X X Body metrics: height All sites X Body metrics Objective motor assessment -Full body wearable wireless intertial sensing-based motion capture system using Xsens MVN to quantify bradykinesia and tremor.-KINARM Endpoint robot coupled with VR-display to extract upper-limb kinematic data.-Zeno walkway to assess gait.(MTC) imaging is performed using a 3D multi-echo gradient echo (GRE) sequence with an MTC pulse to detect and quantify neuromelanin in the substantia nigra and locus coeruleus (He et al., 2023(He et al., , 2021)).Strategically acquired gradient echo, (STAGE) multi-echo, multi-flip-angle susceptibility weighted imaging (SWI), is used to provide the following quantitative data: proton-density, T1-and T2*-maps and simulated double inversion recovery (DIR) as well as SWI minimum-intensity projections and quantitative susceptibility maps (QSM) to quantify the iron in various deep gray matter nuclei (Chen et al., 2018;Liu et al., 2020;Wang et al., 2018). 31phosphorous magnetic resonance spectroscopy (31P-MRS), using a proton-decoupled, nuclear-Overhauser enhanced (NOE) chemical shift imaging (CSI) sequence to assess the intracerebral concentration of phosphorylated metabolites, including nicotinamide adenine dinucleotide (NAD) and adenosine triphosphate (ATP), as previously described (Brakedal et al., 2022).Microvascular imaging is based on modified susceptibility weighted imaging (SWI) sequences, using the ultra-small-superparamagnetic-iron-oxide (USPIO) contrast agent "ferumoxytol" (Buch et al., 2020;Liu et al., 2018).Not all sequences are acquired at all sites and field strengths.

Biosampling
Blood sampling is performed every second year and includes routine clinical tests (see study protocol for complete list) and biobanking of whole ethylenediaminetetraacetic acid (EDTA) blood, serum, PAXgene blood for ribonucleic acid (RNA) analyses, peripheral blood mononuclear cells (PBMCs), platelet isolation for mitochondrial assays, and flash-frozen whole blood samples.The latter is ideal for metabolomic analyses, including the assessment of volatile metabolites, such as ATP and the NAD-metabolome.Cerebrospinal fluid (CSF) sampling is performed every second year.The CSF undergoes routine analyses for cells, protein and glucose, as well as measurement of β-amyloid, total tau protein, and phosphorylated tau protein.Cell-free CSF is biobanked for future analyses.Needle muscle biopsy of the vastus lateralis is performed every second year.Biological samples are stored at the dedicated STRAT-PARK biobank at Haukeland University Hospital.Detailed operating procedures for obtaining and processing each sample type are given in the study protocol and lab manual (supplementary material).

Postmortem examination
Postmortem examination and tissue collection is performed on

Dataset generation
While STRAT-PARK is collecting material for a broad spectrum of future biomarker-related analyses, the following analyses are already planned and will generate data in parallel with cohort recruitment.

Genomics
The genome of all participants will be mapped in DNA extracted from muscle biopsy (or blood where muscle is not available), using wholegenome sequencing (WGS) at mean coverage 30X per base pair (bp) and 150 bp paired-end reads.Muscle DNA is chosen as this tissue is superior to blood for assessing somatic/heteroplasmic mtDNA changes.Cases of monogenic parkinsonism will be identified, and the data stored for future analyses.GBA status will be assessed by targeted sequencing of a preamplified region to ensure specificity, as described (Zampieri et al., 2017).mtDNA will be assessed from the WGS data including heteroplasmic changes at levels as low as 1 %.mtDNA total copy number and proportion of major arc deletion are characterized by quantitative real-time PCR as described (Dölle et al., 2016;Flønes et al., 2018).

Transcriptomics & proteomics
The transcriptome will be assessed by RNA-sequencing in PaxGene whole blood, following ribosomal and globin depletion protocols.The proteome will be assessed in the same blood samples by quantitative proteomics using Tandem Mass Tags (TMT) labeling and mass spectrometry.

Mitochondrial function
The mitochondrial respiratory chain is assessed in PBMCs, muscle, and platelets using a combination of immunohistochemistry (IHC), Western blot (WB), and specific enzyme activities as previously described (Flønes et al., 2018).Serological markers of mitochondrial dysfunction, including fibroblast Growth Factor 21 (FGF21) and Growth/Differentiation Factor-15 (GDF15), are measured in serum and/or CSF using established ELISA assays.FGF21 is a marker of mitochondrial myopathy and is elevated in patients and animals with mitochondrial dysfunction (Lehtonen et al., 2016).Serum GDF15 levels are elevated in patients with mitochondrial disease and respiratory chain deficiencies (Yatsuga et al., 2015).Moreover, increased serum and/or CSF levels of GDF15 in PD have been reported by several studies (Yao et al., 2017).

Markers of proteinopathy and neuronal injury
CSF is undergoing analyses for levels of tau, phosphorylated tau, and amyloid β, as well as α-synuclein seed amplification assays (SAA), to determine the presence of misfolded α-synuclein.SAA was recently reported to show high accuracy in differentiating PD patients from controls.Intriguingly, approximately 7% of the individuals with idiopathic PD were negative for α-synuclein SAA in a large recent study, raising the possibility for stratification (Siderowf et al., 2023).Finally, as a general marker of neuronal injury, neurofilament light chain (NfL) is measured in serum using the Simoa NF-light® assay.

Standardization of data and material collection, storage and management
STRAT-PARK has implemented standard operating procedures across all sites for the collection of data and biological samples.These are described in detail in the study protocol and lab manual (supplementary material).Study personnel receive training in study procedures, including carrying out clinical scale assessment, electronic case report form (eCRF) data registration, and the collection, management, and storage of biospecimens.
Clinical, imaging and molecular data from all centers are registered and stored in a centralized infrastructure, including an eCRF (Viedoc -The New Standard in eClinical Data Management, Viedoc WWW Document (n.d.)), imaging data server (SECTRA PACS for research), and other research servers at Haukeland University Hospital and the University of Bergen).All biological material is stored in the centralized STRAT-PARK biobank organized under The Research Biobank for Aging, Dementia and Neurology at Haukeland University Hospital.

Statistics
Having no indication of the potential number and size of PD biological subtypes, an informative power calculation cannot be performed.Given the clinical heterogeneity and complexity of PD, it is likely that a large population will be required.STRAT-PARK will dynamically adjust the size of the cohort based on results from the first analyses.Descriptive statistics and data analyses for the cohort demographics presented herein were performed using R version 4.2.2 and RStudio 2022.12.0 Build 353.

Ethics
The study is conducted in full accordance with the ICH E6 guideline for Good Clinical Practice and the principle of the Declaration of Helsinki, and the laws and regulations of Norway, including the General Data Protection Regulation (GDPR), and Canada.Ethics committees have approved the study in Norway (REK 74985) and Canada (HSREB 115770 and HSREB 114092).All participants must provide informed consent to be included.

Data sharing
The datasets and code required to reproduce the cohort demographics can be made available upon request to the corresponding author.STRAT-PARK data will be deposited in the European Genome-Phenome Archive (http://ega-archive.org/), a service for permanent archiving and sharing of genetic, phenotypic, and clinical data generated for the purposes of biomedical research.The data will be accessible following an application to the Data Access Committee (DAC).A detailed Data Access Plan is currently being developed and is expected to become available during 2024.

Cohort demographics
We present the preliminary and baseline demographics of the participants.As of March 9th, 2023, 252 participants were included in the STRAT-PARK cohort, comprising 204 individuals with PD and 48 healthy controls.Participation percentage for the different examinations is presented in Fig. 2A.The male to female ratio was 1.8:1 for PD patients and 0.5:1 for controls (Fig. 2B).At baseline, the mean age was 66.2 ± 7.9 and 63.1 ± 10.2 for patients and controls respectively (Fig. 2C).The mean age at diagnosis was 61.9 ± 8.8, while the mean age of onset of motor symptoms was approximately 3.5 years earlier, at 58.6 ± 9.1 years (Fig. 2D).PD patients had a mean disease duration of 7.7 ± 5.4 years (Fig. 3A).At baseline, motor phenotypes were defined according to the TD/PIGD classification.Most patients were classified as either postural instability and gait disorder (N = 91) or tremor-dominant (N = 83), while 20 of the patients were characterized as indeterminate (Fig. 3B).The mean MDS-UPDRS III score was 25.4 ± 12.1, and mean Hoehn & Yahr stage was 2.1 ± 0.6 (Fig. 3C and Table 2).

Discussion
There is an urgent need for identifying biological subtypes of PD to enable precision medicine and care.The aim of the STRAT-PARK initiative is to establish a cohort database and biobank which can be used to conduct bottom-up, biomarker-based classifications of PD, as well as an in-depth characterization of clinicopathological correlates and environmental factors associated with emerging disease subtypes.To this end, STRAT-PARK aligns with recent recommendations on use of molecular biomarkers, deep phenotyping methods, and data-driven analyses (Dorsey et al., 2020;Fereshtehnejad and Postuma, 2017;Hipp et al., 2018;Mestre et al., 2021;Schalkamp et al., 2022), and has several additional advantages, as outlined below.
The biological samples collected, and molecular data generated in STRAT-PARK have been specifically tailored to allow the assessment of biomarkers reflecting principal processes currently associated with the pathogenesis of PD, such as aberrant proteostasis and mitochondrial dysfunction.At the same time, sample and data collection is broad enough to allow new emerging mechanisms and pathways to be explored.
A unique feature of STRAT-PARK is the inclusion of comprehensive biomarkers of mitochondrial integrity, which will be assessed crosssectionally and longitudinally, in successive biospecimens of muscle biopsy and platelets, as well as in the patient brain, using 31P-MRS to assess the levels of phosphorylated energy intermediates and metabolites.Furthermore, micro-vascular imaging with ferumoxytol will detect and quantify micro-angiogenic changes which may be associated with mitochondrial dysfunction, similar to that seen in mitochondrial disorders (Reichard and Asosingh, 2019), and/or neuroinflammation (Desai Bradaric et al., 2012).In vivo assessment of proteinopathy markers will allow the cohort to be stratified based on the composition of proteinopathy, and its longitudinal change in time.This will also allow us to further explore proposed associations between mitochondrial impairment, neuroinflammation and proteinopathy (Flønes et al., 2022;Flønes and Tzoulis, 2022).Post-mortem neuropathological examination will provide the ground truth against which mitochondrial and proteinopathy biomarkers will be validated.
Finally, wearable sensor technology will provide quantitative measures of bradykinesia, tremor, mobility, dyskinesia, gait and axial motor impairments in PD, that are objective and may be more sensitive than clinical scale-based motor assessment (Memar et al., 2018).These assessments will provide multiple high-resolution measures of motor function, which are ideal for data-driven stratification.Moreover, wearable sensors will enable monitoring of physical activity and exercise, which may play an important role in influencing clinical and molecular measures, as well as disease progression (Tsukita et al., 2022).
An additional advantage of the STRAT-PARK initiative is that, thanks to the diverse clinical and molecular measures it provides, it can support recently proposed frameworks for biological classification of PD, including NSD-ISS and SynNeurGe (Höglinger et al., 2024;Simuni et al., 2024).
So far, the STRAT-PARK project has shown satisfactory progress and is well underway to becoming a large international PD cohort.With the current recruitment rate, we expect the cohort to be fully included by 2035.However, data are being continuously curated and the dataset from the first 300 participants is expected to be analysis-ready in June 2024.As expected, healthy controls are more challenging to recruit than patients, and are currently in a ratio of 1:4, rather that the desired 1:3.The sex balance of the cohort is similar to that reported from other cohorts (Marek et al., 2018;Mollenhauer et al., 2013).The cohort is heterogenous in terms of severity of both motor and non-motor symptoms, illustrated, for instance, by substantial variation in the MDS-UPDRS III and MDS-NMS total scores.
Currently, the STRAT-PARK cohort has an overrepresentation of Hoehn & Yahr stage 2, although the range is 0 -4.As the study progress, we anticipate the distribution of severity stages to become more even, as more newly diagnosed patients are being recruited, and included patients progress.To encourage participation across disease stages, the protocol offers some flexibility, such as declining or withdrawing from certain procedures (i.e.muscle biopsy, lumbar puncture, and MRI).Furthermore, at the Norwegian sites all clinical assessment is conducted in the "ON" state, which may be of benefit to patients with advanced disease.The study has an active patient and public involvement (PPI) component and collaborated closely with patient associations and family physicians' networks to reach patients of all disease stages.
The STRAT-PARK study has several limitations.Cognitive function is only tested using MoCA, which makes it difficult to thoroughly characterize the cognitive impairment of the participants.Olfaction is evaluated using B-SIT, which is currently not used in any of the largest PD cohorts (e.g.Parkinson Progression Marker Initiative (PPMI) uses UPSIT (Marek et al., 2011) and Oxford Parkinson Disease Center (OPDC) discovery cohort uses Sniffin' Sticks (Szewczyk-Krolikowski et al., 2014)).Additionally, the sensitivity to diagnose olfactory dysfunction is lower for B-SIT than UPSIT (Doty et al., 1996).However, it has the advantage of being easy and quick to administer, valid in cross-cultural settings and it is possible to convert UPSIT scores to B-SIT scores (Lawton et al., 2016).The fact that patients are primarily tested in the "ON" state minimize discomfort, but also presents challenges in assessing the impact of dopaminergic treatment on the MDS-UPDRS III score.Nonetheless, evaluating motor function in the "ON" state is likely to provide a more accurate reflection of how patients are typically monitored in real-world scenarios.This may be of value when translating our findings to clinical practice.
In conclusion, currently, there is no consensus on how to subtype PD (Albrecht et al., 2022) and most existing classification systems rely on clinical features alone, which provide no insight into molecular pathogenesis and therapeutic targets.Achieving molecular stratification of PD could change our understanding of the disease, from that of a highly heterogeneous clinically defined syndrome of obscure etiopathogenesis, to specific disease subtypes, each driven by defined molecular processes.Accounting for such subtypes would decrease the complexity and heterogeneity of the PD spectrum and allow research to focus on pathophysiologically more homogeneous samples and/or cohorts, thereby increasing the signal-to-noise ratio in both mechanistic and treatment studies.Biomarkers of mechanistically defined PD subtypes would allow for selecting patients for trials of targeted therapy, ultimately increasing the likelihood of positive outcome.
participants who have consented to this.Autopsy is performed within 48 hours after death.The left cerebrum, cerebellum and brain stem are divided into anatomical regions, snap-frozen in liquid nitrogen and stored at − 80 C. The right cerebrum, cerebellum and brain stem are fixed in formalin, divided into anatomical areas and embedded in paraffin blocks.In addition, samples are collected for formalin fixation and freezing from the heart, muscle, liver, kidney, and each anatomical segment of the intestine.Postmortem tissue samples are stored at the dedicated STRAT-PARK biobank at Haukeland University Hospital.

Fig. 1 .
Fig. 1.Overview figure of the STRAT-PARK initiative.Data from clinical assessment and questionnaires, imaging, neurophysiological examinations, biosampling and postmortem examinations will ultimately be used to employ multidimensional data integration and data-driven analyses to identify and validate stratification biomarkers for PD.Created with BioRender.com.

Fig. 2 .
Fig. 2. Demographic data extracted from the baseline visit of the participants in the STRAT-PARK cohort.(A) Bar plot of percentage of participants who have successfully undergone the different examinations.The examinations listed (except for EEG and OMA) are currently being performed at all three sites."Routine blood" is routine blood samples, and "Biobank blood" is blood samples collected for the biobank.Abbreviations: EEG, Electroencephalogram; OMA, Objective Motor Assessment; MRI, Magnetic Resonance Imaging; CSF, Cerebrospinal Fluid.(B) Bar plot of sex-distribution of the participants.(C) Violin plot of age at baseline (years) for patients and controls.(D) Violin plot of age at diagnosis and age at symptom onset (i.e., age at onset of motor symptoms).

Table 1
Currently performed examinations and biosampling.

Table 2
Cohort characteristics.: B-SIT, Brief Smell Identification Test; MDS-UPDRS, International Parkinson and Movement Disorder Society Unified Parkinson's Disease Rating Scale; MDS-NMS, International Parkinson and Movement Disorder Society Non-Motor Rating Scale; MoCA, Montreal Cognitive Assessment. Abbreviations