Cognitive Heterogeneity in Parkinson’s Disease: A Mechanistic View

Summary

Cognitive impairment occurs in the vast majority of individuals with Parkinson’s disease (PD), exacting a high toll on patients, their caregivers, and the health care system. In this review, we begin by summarizing the current clinical landscape surrounding cognition in PD. We then discuss how cognitive impairment and dementia may develop in PD, based on the spread of the pathological protein alpha-synuclein (aSyn) from neurons in brainstem regions to those in the cortical regions of the brain responsible for higher cognitive functions, as first proposed in the Braak hypothesis. We appraise the Braak hypothesis from molecular (conformations of aSyn), cell biological (cell-to-cell spread of pathological aSyn), and organ-level (region-to-region spread of aSyn pathology at the whole brain level) viewpoints. Finally, we argue that individual host factors may be the most poorly-understood aspect of this pathological process, accounting for substantial heterogeneity in the pattern and pace of cognitive decline in PD.

In brief

Carceles-Cordon et al. discuss molecular and cellular mechanisms of alpha-synuclein spread in the brain and examine their interplay with host factors such as genetic background to provide an overview of the pathogenesis of cognitive heterogeneity in Parkinson’s disease.

Introduction

Parkinson’s disease (PD) and the related disorder dementia with Lewy bodies (DLB) are defined by deposition of misfolded alpha-synuclein (aSyn) into neuropathological inclusions known as Lewy bodies (LBs) and Lewy neurites (LNs) in neurons, and both are considered Lewy body disorders (LBD). Indeed, at autopsy, brain samples from people who bore a clinical diagnosis of PD and those who bore a diagnosis of DLB may be virtually indistinguishable. During life, however, individuals with PD and DLB patients manifest very differently, with DLB patients demonstrating cognitive symptoms and dementia from the onset of their disease, whereas PD patients initially show motor symptoms of bradykinesia, rigidity, tremor, and postural instability.¹ In addition, while most people with PD eventually develop cognitive decline and dementia, a minority do not – and even among those PD patients who develop dementia (Parkinson’s disease with dementia, or PDD), the pace varies widely.^2–4 Indeed, the time scale over which PDD may develop ranges from months, to years, to decades.

This clinical heterogeneity is a critical factor in disease management. An individual presenting with isolated motor symptoms and newly diagnosed with PD will have different needs from an individual diagnosed with PDD, who may require social or family support to perform activities of daily living. Indeed, cognitive decline in PD has been shown to lead to decreased quality of life for the patient, higher burden for the caregiver, and increased cost of care for the health care system.^5–7 Moreover, the number of people affected by PD is growing, projected to exceed 12 million worldwide by 2040.⁸ Because this increase is largely driven by population aging, and because age is strongly associated with cognitive impairment in PD, the projected increase in PD prevalence will almost certainly come with a substantial increase in PDD.

Here, we first summarize the clinical landscape of cognitive heterogeneity in PD, providing an overview of current treatments for cognitive impairment in PD, as well as experimental approaches in clinical trials. We then discuss how PDD may develop, based on the spread of aSyn pathology from brainstem regions to higher-order cortical regions of the brain, as first proposed in the Braak hypothesis. We appraise the Braak hypothesis from molecular (conformations of aSyn), cell biological (cell-to-cell spread of pathological aSyn), and organ-level (region-to-region spread of aSyn pathology at the whole brain level) viewpoints, pointing out areas that are more vs. less well-supported by existing data. Finally, we suggest that host factors – which we define as characteristics of individuals that influence susceptibility and temporal course of a disease, such as genetic background or environmental exposures – may be the most poorly-understood aspect of cognitive heterogeneity in PD and, as such, represent a particularly valuable area for investigation.

Epidemiology, Presentation, and Assessment of Cognitive Symptoms in Parkinson’s Disease

Between 15-35% of PD patients will meet criteria for mild cognitive impairment (PD-MCI) at the time of diagnosis,⁹ and several longitudinal studies have demonstrated that approximately 80% of patients will eventually progress to dementia (PDD).¹⁰ More recently, changes in cognition have even been reported prior to PD diagnosis¹¹ in population-based studies.^12,13 Additionally, prospective studies have demonstrated cognitive worsening in individuals with prodromal or at-risk PD, such as those having impaired olfaction and dopamine transporter (DAT) deficits¹⁴ or REM sleep behavior disorder (RBD).¹⁵

Traditionally, primary cognitive deficits in PD were attributed to frontal lobe-based changes (manifesting as difficulty with working memory, executive ability and attention).¹⁶ However, recent work has found that initial impairments can occur in a range of cognitive domains, including executive, memory, visuospatial, attentional, and even language domains.^17–19 Moreover, additional neuropsychiatric symptoms that can either impair cognition or are often comorbid with cognitive impairment in PD include depression, apathy, and psychosis (i.e. hallucinations and delusions). Clinical predictors of cognitive impairment include higher current age, higher age at disease onset, greater disease severity, the postural instability-gait disorder subtype, and presence of features such as RBD, psychosis, depression, and anxiety.²⁰

Advancements in the diagnosis and assessment of cognition in PD include the establishment of diagnostic criteria for both PDD²¹ and PD-MCI.²² PDD is characterized by cognitive decline over time, cognitive impairment in multiple domains on testing, and significant cognitive functional impairment.²¹ PD-MCI consists of self- or other-reported cognitive decline, with some cognitive impairment on testing, but no significant cognitive functional impairment (i.e. preserved instrumental activities of daily living). Other improvements in the diagnosis and evaluation of cognitive symptoms in PD are summarized in several recent reviews, which cover recommended cognitive testing,²² global cognitive assessment instruments for use in PD,²³ and comparisons of the relative sensitivity of commonly-used cognitive tests in non-demented PD patients.²⁴ Additional recommendations for assessment of cognition in PD can also be found through the National Institute of Neurological Disorders and Stroke Common Data Elements (https://www.commondataelements.ninds.nih.gov).

Dementia with Lewy bodies

A point of controversy related to cognition in PD pertains to the related disorder of DLB, and whether PDD and DLB are categorically, or simply dimensionally, different.⁷ DLB is characterized by dementia at time of diagnosis and some combination of parkinsonism, psychosis, and RBD. DLB is similar neuropathologically to PDD, particularly in terms of Lewy body pathology, although some have argued for differences between PDD and DLB with respect to (1) the degree of concomitant Alzheimer’s disease (AD) pathology, which may be higher in DLB; (2) biomarker profiles in DLB versus PDD;^25–27 and (3) possible genetic variability in SNCA (encoding aSyn, the hallmark protein in the pathological lesions of PD, as described below).²⁸ In practice, a controversial “one-year rule” is used, with dementia that precedes or occurs within one year of onset of parkinsonism being labeled as DLB and all other cases being diagnosed as PDD.¹

Management of Cognitive Impairment in PD – Current and Future Prospects

Management options for cognitive impairment in PD consist largely of avoiding medication side effects and treating co-existing conditions that may worsen cognition. For example, medications with anticholinergic properties are sometimes used in PD,²⁹ but they are associated with long-term cognitive decline³⁰ and should be avoided in older individuals with PD. Comorbid vascular diseases, other neuropsychiatric symptoms, obstructive sleep apnea and other sleep disorders,³¹ and orthostatic hypotension, all associated with worsening of cognitive impairment in PD, should be treated.

To improve cognition in PD, there is evidence that acetylcholinesterase inhibitor (ChEI) treatment is of modest benefit,^32,33 with lesser effects reported for memantine.^34–36 Currently, no medications have been approved for the symptomatic management of PD-MCI, with negative randomized controlled trials (RCTs) for rasagiline (MAO-B inhibitor),³⁷ atomoxetine (selective noradrenaline reuptake inhibitor),³⁸ and rivastigmine (ChEI).³⁹ Other compounds recently or currently being tested for PD cognitive impairment include SYN120 and itepridine (5-HT₆ antagonists), mevidalen (selective positive allosteric modular [PAM] of the D1 receptor),⁴⁰ blarcamesine (intracellular sigma-1 receptor agonist), IRL752 (a small molecule that selectively enhances norepinephrine, dopamine, and acetylcholine neurotransmission in the cerebral cortex),⁴¹ and CST-103 (beta-adrenoreceptor agonist) targeting the locus coeruleus to improve noradrenergic tone. There may also be beneficial effects of physical exercise and cognitive training on cognitive functions in PD.^42–44

Taken together, however, current options for treating cognitive impairment and dementia in PD are limited.⁴⁵ Most therapies aiming to stabilize or reverse the underlying disease process leading to cognitive decline in PD only exist in clinical trials. Examples of these experimental therapies under investigation now include the MAP kinase inhibitor neflamapimod, currently being tested in DLB (NCT04001517); the antibiotic ceftriaxone, purported to reduce glutamatergic hyperactivity and subsequent excitotoxicity; a proprietary plasma fraction (GRF6021) that aims to enhance neurogenesis and reduce neuroinflammation; two N-methyl-D-aspartate (NMDA) receptor PAMs (SAGE-718 and NYX-458); and ambroxol (repurposed expectorant/mucolytic drug, aiming to increase levels of the lysosomal enzyme glucocerebrosidase in order to reduce aSyn aggregation).

Despite the current scarcity of treatment options, however, increasing evidence from prospective studies examining newly diagnosed patients⁴⁶ and those with prodromal conditions, such as RBD,⁴⁷ suggest that a therapeutic disease-modifying window exists where progression to significant cognitive impairment could be halted or delayed. Thus, this is an appropriate time to focus research efforts on better understanding the processes leading to cognitive decline in PD, in order to interfere in these processes therapeutically.

Neuropathological substrates for cognitive impairment in PD

The neuropathological hallmark of PD is the neuronal LB, a dense, cytoplasmic inclusion body found at autopsy in individuals who were diagnosed with PD during life. LBs were first described over 100 years ago and are found throughout the PD brain, in areas that are susceptible to neurodegeneration.⁴⁸ The main protein within LBs is aSyn,⁴⁹ a normally-occurring protein whose causal role in the development of PD is supported by the fact that both missense mutations in⁵⁰ and increased copy numbers of⁵¹ its coding gene SNCA result in familial PD and DLB. The normal function of aSyn is still unclear, although strong evidence exists for a role in synaptic vesicle function,^52,53 where aSyn may exist in equilibrium between cytosolic, unfolded forms and membrane-bound, helical forms.⁵⁴ In disease contexts, aSyn becomes hyper-phosphorylated, particularly at serine 129,^55,56 and aggregates into more insoluble forms.⁵⁷ In addition, Lewy neurites (LNs), abnormal neuronal processes also containing pathological forms of aSyn, are found in both DLB and PD.

LBs and LNs are found across many brain regions, including areas that exhibit selective vulnerability such as the dopaminergic neurons of the substantia nigra. Moreover, across many neuropathological cohorts, Lewy pathology appears to follow a pattern of regional involvement, leading to the development of a commonly-used staging system by Braak and colleagues.⁵⁸ In brief, Braak staging proposes that Lewy pathology initially occurs in the dorsal motor nucleus of the vagus (dmX) and the anterior olfactory nucleus (stage I); followed by involvement of the raphe nuclei and the gigantocellular reticular nucleus of the medulla oblongata (stage II). Subsequent stages involve brainstem areas, including the midbrain dopaminergic neurons of the substantia nigra pars compacta (stage III), as well as prosencephalic regions and the anteromedial temporal cortex (stage IV). This rostro-caudal spread culminates in the presence of LBs in the neocortex (stages V and VI).

While this review will concentrate on PD, PDD, and the related condition DLB, we note that aSyn pathology may also occur in other cell types, such as oligodendrocytes. These aSyn glial cytoplasmic inclusions (GCI), found predominantly in oligodendrocytes, characterize multiple system atrophy (MSA), a disorder presenting with parkinsonism and ataxia, to varying degrees.⁵⁹ Interestingly, MSA patients are less likely to develop cognitive symptoms than PD patients, to the extent that development of dementia was once considered grounds for re-evaluation of an MSA diagnosis,⁶⁰ despite an overall faster rate of motor progression.

An under-recognized neuropathological feature of PD, and particularly PDD and DLB, is the frequent co-existence of the amyloid-beta-containing plaques and tau-containing neurofibrillary tangles of Alzheimer’s disease (AD).⁶¹ While a smaller proportion of PD/PDD patients (~25-30%) have extensive neocortical tangle pathology, and a larger proportion of patients have concomitant amyloid plaque pathology,^62,63 approximately half of PD patients demonstrate enough concomitant plaque and tangle pathology to warrant a secondary neuropathological diagnosis of AD.^61,64,65 Furthermore, concomitant AD pathology may be even more prominent in DLB than in PD and PDD.^26,66

Intriguingly, multiple lines of evidence suggest some degree of synergy between aSyn pathology and AD pathology, and particularly tau pathology. Among human neuropathological studies comparing brain regions in cases with confirmed LBs upon autopsy, the severity of AD pathology appears to correlate with the aSyn pathology burden in both PD and DLB.^67,68 Moreover, in vitro studies have long demonstrated that aSyn induces the fibrillization of tau and that co-incubation of tau and aSyn increases the fibrillization of both pathological proteins.⁶⁹ More recently, Bassil et al. have reported in vivo evidence for synergy between aSyn and tau, demonstrating that co-inoculation of pathological forms of aSyn and tau into mouse brains accelerates the development of tau pathology.⁷⁰

While neuropathological studies are, by necessity, static “snapshots” of a single time point per individual in a years-long, or even decades-long, process, it is worth considering the inter-individual differences in tempo among patients with LB disorders in both the onset and the progression of cognitive decline. That is, functional neuroanatomy would suggest that brainstem-only-pathology cases would be unlikely to have prominent cognitive deficits, and this is in fact supported by the clinico-pathological literature.^71,72 An open question, however, is why some individuals go for years, or decades (or their entire disease course), without development of cognitive decline/cortical pathology, while others develop cognitive deficits/cortical pathology as a presenting symptom.

The Braak hypothesis: A model for the development of cognitive decline in PD?

Current understanding of the pathogenesis of cognitive dysfunction in PD is largely informed by the seminal studies of Braak and colleagues.^58,73,74 Braak’s original study consisted of 110 patients with LB inclusions who were stratified into six topographically distinct, progressive stages of PD-related brain lesions,⁵⁸ wherein dysfunction is posited to be a direct consequence of the neurotoxic effect of misfolded aSyn accumulating in LBs.⁷⁵ Indeed, aSyn-containing LBs are found in neocortical areas and other brain regions required for higher cognitive functioning,⁷⁶ and their presence correlates with neuronal cell death and dementia. In the diffuse neocortical stage proposed by Braak (stages V and VI), characterized by widespread aSyn inclusions in the brain, prevalence of dementia has been reported to be nearly universal.^58,75

In parallel with this staging system, Braak’s studies proposed that these neuropathological stages are the traces of a disease process whereby pathological species of aSyn spread within the CNS in a rostro-caudal manner, propagating from the lower brain regions to cortical areas where they then perturb normal cognitive processes. Specifically, Braak and colleagues speculated that a neurotropic pathogen could reach the CNS through the nose (anterograde ascension) or the enteric system (retrograde ascension via the enteric plexus and preganglionic fibers of the vagal nerve).⁷⁷ This pathogen, which was initially speculated to be a viral agent due to other classical examples of rostro-caudal ascension by other viruses,⁷⁸ would potentially reach the medulla and from there it would spread to the midbrain and above.⁷⁷ In support of nasal or vagal nerve routes of entry, aSyn inclusions can be found in the enteric nervous system,^79–81 as well as in the submandibullary gland,⁸² and the anterior olfactory bulb^83,84 of PD patients. Moreover, clinical symptoms suggesting early involvement of the olfactory system, the vagal nerve, and/or the gut are present in PD. Specifically, hyposmia and dysautonomia may precede parkinsonism in the prodromal stage of PD,^85–87 and constipation is highly prevalent among those newly diagnosed with PD, often preceding overt development of parkinsonism.⁸⁸

Over the last 15 years, no clear evidence of a viral pathogen has emerged. However, evidence suggesting that aSyn pathology might propagate within the brain has steadily accumulated. Early signs came from two independent studies of fetal mesencephalic dopaminergic neuronal grafts in PD patient brains examined by autopsy years after graft implantation. Surprisingly, LB pathology was observed in grafted cells.^89,90 Because the grafts came from fetal tissue which would be pathology-naïve at implantation, these studies strongly suggested propagation of aSyn pathology from the surrounding host tissue. Experimental evidence for this possibility emerged shortly thereafter, with the demonstration that neurons can take up and transmit aSyn to other cells,⁹¹ and that exogenous preformed fibrils (PFFs) constituted from recombinant aSyn could seed intracytoplasmic aggregation of endogenous aSyn in neuronal culture.^92,93 This in cellulo work was followed by a report that a single intracerebral injection of aSyn PFFs in the dorsal striatum of both wild-type and aSyn-transgenic mice could recapitulate the in vivo spread of aSyn pathology to many areas of the brain, in a time-dependent manner, with subsequent neurodegeneration and development of motor and cognitive deficits.^94,95 These key findings have subsequently been replicated widely.^96–100 Moreover, studies applying this PFF-injection model to non-human primates have confirmed aSyn pathological spread far from the injection site, accompanied by neurodegeneration in those brain regions affected by the pathological spread.^101,102 More recently, aSyn pathological spread from an initial site of injection has been demonstrated for human brain-derived pathological material,¹⁰³ which may be more potent than PFFs generated from recombinant aSyn,¹⁰⁴ on a variety of mouse model backgrounds.^97,105 In addition, injection of aSyn PFFs into the stomach and small intestine has been reported to result in the subsequent development of brain pathology in wild-type mice and rats,^106,107 although in other studies using similar models the development of brain lesions has been transient¹⁰⁸ or restricted to the dmX.¹⁰⁹

The stereotypical patterns of Lewy pathology demonstrated by Braak in human brain tissue,⁵⁸ together with evidence for cell-to-cell transmission of aSyn in cellular and mouse models, in the absence of evidence for an exogenous viral agent, has led to the current prevailing view: that misfolded aSyn itself acts as the neurotropic pathogen responsible for spreading pathology across the CNS in a prion-like manner.^110–112 Specifically, pathologically-misfolded aSyn might serve as the template for subsequent misfolding of aSyn in a pathology-naïve cell in order to convert aSyn in the recipient cell into a pathological form. In support of this view is the dependence of aSyn spread models on the existence of endogenous non-pathological aSyn in the host, because injection of PFFs in alpha-synuclein knockout mice causes no pathology.⁹⁴

Evidence for cell-to-cell transmission of alpha-synuclein as the mechanism underlying cognitive decline in PD

For a prion-like version of the Braak hypothesis to account for cognitive decline in PD, several conditions must be met. First, pathological forms of aSyn should be able to transfer from affected cells to pathology-naïve cells. Second, pathological aSyn from the affected cells should then cause templated misfolding of endogenous aSyn in the naive cells into pathological forms. Third, pathological aSyn should cause neurodegeneration and cognitive decline. Fourth, at the human brain level, there should be an orderly progression of pathological involvement of rostral-to-caudal brain regions. We take each of these conditions in turn and examine existing evidence (Figure 1).

Figure 1. Appraising the Braak hypothesis as the mechanistic basis for cognitive decline in Parkinson’s Disease.

The literature supports cell-to-cell transfer of aSyn from affected to pathology-naïve cells and the subsequent ability of pathologic aSyn to cause templated misfolding of endogenous aSyn in a prion-like manner. However, disconnects emerge with respect to whether aSyn pathology always forms the basis for neuronal dysfunction and cognitive impairment, and whether aSyn pathology consistently follows a rostro-to-caudal spread pattern in the human brain.

Pathological forms of aSyn should be able to transfer from affected to pathology-naïve cells

For pathological aSyn to be able to transfer from affected to pathology-naïve cells, it must exist in an available space, and mechanisms must exist for uptake into recipient cells, including neurons that will ultimately degenerate. Despite lacking a secretory signaling peptide sequence and localizing to the cytosol, both monomeric and aggregated forms of aSyn are found in the extracellular space,^113,114 and aSyn is found in the CSF, plasma, and serum of both healthy individuals and PD patients.^115,116 Non-classical secretory pathways, including both passive and active mechanisms, are thought to be involved in the release of aSyn from cells into the extracellular space.¹¹⁷ A fraction of the secreted aSyn fraction has been linked to exosomes, extracellular vesicles originating from the endosomal compartment,^118,119 which have been reported to induce Lewy pathology in vivo.¹²⁰ Specifically, exosomes harvested through ultracentrifugation of DLB brain tissue induced intracellular and dendritic aSyn aggregation in neurons and astrocytes of WT mice.¹²⁰ Moreover, exosomal aSyn is increased in PD patients when compared to controls and has been proposed as a biomarker of PD severity and a blood-based diagnostic test.^121–123

Neuronal uptake of amyloid aggregates of aSyn occurs at the somatodendritic compartment as well as at the axon and presynaptic terminals through the endosomal pathway.^124–126 Several cell surface proteins have been proposed as receptors that facilitate internalization of aSyn into the cell.¹²⁷ For example, lymphocyte activation gene 3 (LAG3) reportedly binds to PFFs and mediates internalization of aSyn,¹²⁸ although these findings have been questioned.¹²⁹ Recent evidence also suggests that cellular prion protein (PrP^C) mediates uptake and toxicity of aSyn¹³⁰ and that physical interaction between aSyn fibrils and PrP^C can mediate cognitive dysfunction through alteration of hippocampal long-term potentiation.¹³¹ Once internalized, aSyn can be transported in both anterograde and retrograde directions along neuronal processes;^125,132,133 however, it is still unclear how fibrils escape from endocytic vesicles and trigger intracellular pathology.

Recently, we have demonstrated that the PD risk-associated gene GPNMB, which encodes glycoprotein non-metastatic melanoma protein B (GPNMB), may be involved in the cell-to-cell transfer of pathological aSyn.¹³⁴ Specifically, we have shown that decreasing GPNMB levels in human iPSC-derived cortical neurons results in greatly reduced uptake of exogenously-added aSyn PFFs, with subsequent reduction of pathology (i.e. insoluble, hyperphosphorylated aSyn aggregates) in the recipient cells. Of interest, GPNMB is a transmembrane protein expressed on the cell surface which can be cleaved to generate a soluble form, and addition of GPNMB is sufficient to increase uptake of aSyn PFFs in cell types that normally do not express GPNMB highly. GPNMB and aSyn co-localize and co-immunoprecipitate, suggesting that the interaction between these proteins may be responsible for these cellular effects. Finally, higher GPNMB levels associate with PD risk-associated genotypes in human CSF and brain, and higher GPNMB levels are found in PD patient plasma (compared to neurologically-normal controls), with the highest levels in the most severely affected patients.

Pathological aSyn from affected cells should cause templated misfolding of naive aSyn into pathological forms

Evidence for templated misfolding of aSyn has been reported at in vitro, in cellulo, and in vivo levels. In the in vitro context, protein misfolding cyclic amplification (PMCA) and real-time quaking-induced conversion (RT-QuIC) are two diagnostic assays developed in recent years that take advantage of the self-templating capacity of misfolded proteins.^135–137 In aSyn PMCA and RT-QuIC assays, adapted from similar assays developed in the prion field, samples derived from patient biofluids that contain aSyn “seeds” (presumably possessing a particular conformation) are first incubated with unfolded recombinant aSyn protein substrates. These “seeds” serve as templates for the folding of substrate aSyn into a form that can bind amyloid dyes. Through multiple rounds of templating, followed by mechanical fragmentation of the resulting amyloid fibrils, an exponential amplification is achieved. The kinetics of this amplification can then be quantified using amyloid-binding dyes. The finding that both of these assays can distinguish PD from controls with high sensitivity and specificity supports the idea that CSF from PD patients contains “strains” (i.e. specific conformations) of aSyn that can be propagated.^138–140 We note, in particular, that the aSyn-PMCA assay developed by Shahnawaz, Soto, and colleagues has been reported to diagnose PD patients with >88% sensitivity and >96% specificity based on their CSF samples.¹³⁹ Moreover, CSF PMCA assays detecting aSyn oligomers have been reported, in conjunction with enzyme-linked immunoassay measures of CSF neurofilament light chain (NFL), to discriminate between early MSA and LBD.¹⁴¹

Distinct aSyn strains, defined as forms of aSyn with differential seeding capacity and pathogenic behavior, can be generated in cells. For example, Guo, V. Lee, and colleagues discovered two synthetic aSyn PFF strains with differential ability to recruit tau into pathological inclusions.¹⁴² In parallel, another group also characterized two different morphologies of aSyn pathology, fibrils and ribbons,¹⁴³ reporting that when inoculated into rats, each type of aSyn conformation produced a distinct histopathological and behavioral phenotype.¹⁴⁴ In addition, aSyn GCI derived from human MSA brain appear to be conformationally and behaviorally distinct from aSyn strains found in LBs in PD.^103,145 Indeed, GCI-derived aSyn demonstrated approximately 1000-fold more seeding capacity than LB-derived aSyn in cultured oligodendrocytes, in cultured neurons, and in WT mouse brain. More recently, Yang and colleagues reported, using cryo-electron microscopy, that the ultrastructure of aSyn filaments derived from MSA patients differs markedly from aSyn filaments derived from PD, PDD, and DLB patients.¹⁴⁶ Thus, ultrastructural studies, work in cell and rodent models, and human biomarker studies all support the existence of distinct “strains” of aSyn (Figure 2). We note that while aSyn does not exhibit the infectivity of prion proteins, the concept of “strains,” borrowed from the prion literature, is relevant here. Specifically, the prion field has recognized for many years that the same protein – PrP^c – underlies different clinical manifestations in humans, with clinical syndromes resulting from the biological behavior of the misfolded prion protein, referred to as PrP^sc (reviewed in Aguzzi et al.,¹⁴⁷ and Sigurdson et al.¹⁴⁸). Not all conformations of PrP^sc are the same, though, and these distinct conformations manifest as distinct, yet reproducible, syndromes in affected individuals, resulting in the “strain” concept. Thus, the accumulating evidence for distinct conformations of aSyn, with different biological properties, supports the hypothesis that aSyn conformations may dictate “strains” of aSyn, which in turn contribute to the cognitive heterogeneity seen in the LBD.

Figure 2. Evidence for distinct aSyn “strains” in the pathogenesis of synucleinopathies.

Cryo-EM studies support the existence of distinct conformations of aSyn in MSA vs. PD, both characterized by the presence of aSyn pathology. Studies of distinct conformations of aSyn, or aSyn species isolated from MSA vs. PD brain, demonstrate differential seeding and spreading properties in cell and animal models. CSF samples from individuals with MSA vs. PD exhibit distinct fibrillization properties in in vitro amplification assays.

Dysfunction of clearance mechanisms within the cell likely contributes to both cell-to-cell transmission of pathological forms of aSyn and opportunity for pathological forms of aSyn to template the misfolding of native aSyn. In support of this point, antibody-mediated clearance of aSyn reduces extracellular levels and decreases intercellular transmission in a transgenic mouse PD model;¹⁴⁹ whereas inhibition of clearance mechanisms results in the opposite effect.¹⁵⁰ Notably, dysfunction of autophagosome-lysosome pathways responsible for clearance of aggregates has been demonstrated in PD, contributing to worsening of aSyn aggregation,^151–156 and lysosomal pathway functions are enriched among PD risk genes.^157–159 In this context, we highlight the work of Mazzuli, Krainc, and colleagues, demonstrating that dysfunction of the lysosomal enzyme glucocerebrosidase (encoded by the PD-associated risk gene GBA) results in increased aSyn aggregate formation in neurons.¹⁶⁰ These aggregates in turn accumulate in lysosomes, causing greater impairment in glucocerebrosidase function in a “feed-forward” loop. Recently, the same group has demonstrated that directly boosting wild-type glucocerebrosidase activity may reduce pathogenic phenotypes in patient-derived dopaminergic neurons.¹⁶¹

“Pathological aSyn” should cause neurodegeneration and cognitive decline

While experimental evidence considered so far has been largely supportive of aSyn as a prion-like protein, disconnects do emerge when we consider the evidence for this prion-like activity as a mechanism for development of cognitive decline and dementia. For example, the connection between cortical aSyn pathology and cognitive decline and dementia is not straightforward.¹⁶² Incidental LBs (iLBs) are found in approximately 8-18% of neurologically normal individuals on post-mortem examination, including in higher-order cognitive areas.^163,164 The accumulation of iLBs may correlate with age: widespread aSyn iLBs without cognitive impairment can often be observed among centenarians.¹⁶⁵ Notably, in one of the largest iLB pathology cohorts reported to date, 55% of subjects who would have been classified as Braak stages V and VI (diffuse neocortical involvement with predicted dementia) lacked signs of dementia or extrapyramidal symptoms.¹⁶⁶ According to Braak’s original cohorts, cognitive decline would be expected with an increasing prevalence from stage III (36% with cognition affected) to stage VI (100% with cognition affected).⁷⁵

At the same time, a number of disease-focused neuropathological studies have found that most PD cases are compatible with the Braak staging system,^167–170 and Braak staging has proven particularly useful in early-onset and prolonged-duration PD.^167,171 Importantly, when only subjects with a clinical diagnosis of dementia were considered, correlation with a high Braak stage was excellent.¹⁶⁶ These seemingly-contradictory results may be reconciled if LBs do not necessarily result in cell dysfunction and death, as evidenced by the presence of iLBs in non-diseased subjects. Rather, LBs could be required, but not sufficient, for cell death,¹⁷² accounting for their high prevalence in autopsy series of PD and DLB cases. Alternately, some have argued that mature LBs could play a protective role in the brain by accumulating increasing quantities of aSyn oligomers in vulnerable individuals and thus preventing their neurotoxic effects.¹⁷³

Cortical pathology in the human brain should follow a rostral-to-caudal pattern of spread

While the most prevalent pattern of LB pathology appears to follow Braak staging, it is worth noting that many cases do not follow this pattern. That is, collective evidence from multiple studies suggests that approximately 7-29% of LBD confirmed diagnoses do not show LBs in brainstem structures,^174–177 arguing that rostro-caudal spread of aSyn is not the only route of PD progression. Moreover, correlation between the severity of LBs in lower brainstem regions and severity of LB pathology in the neocortex is poor,¹⁷⁷ suggesting that factors other than abundance of LB pathology in lower-order brain regions govern which individual cases go on to develop cortical pathology. Indeed, Surmeier, Obeso, and Halliday have argued that the pattern of aSyn pathology found in human studies, connectome-mapping studies, and the Braak hypothesis cannot be fully reconciled, also suggesting the existence of other factors that “gate” or regulate the pace of pathological aSyn spread in the brain.¹⁷⁸ In contrast, in mouse models in which aSyn fibrils have been introduced by a single intrastriatal injection, neuronal connectivity, together with aSyn expression, explains a substantial proportion of the resultant aSyn pathology observed throughout the brain at subsequent time points.^98,100 Taken together, these data suggest that variability in humans, in the pace and pattern of aSyn pathological spread to those areas of the brain responsible for cognition, is far greater than in model systems. Moreover, neither the Braak hypothesis, nor neuroanatomical connectivity, can adequately account for this observed variability.

Individuals “host” differences: Impact on mechanisms of cognitive decline in PD

If LB are required, but not sufficient, for degeneration of cortical neurons and cognitive decline, how might we then explain the heterogeneous cognitive outcomes observed in subjects with abundant aSyn pathology? Moreover, if neither the Braak hypothesis nor current knowledge of the human brain connectome can fully explain whether aSyn pathology spreads from the brainstem to the limbic and neocortical areas in a given individual, what, then, might account for this variability? We argue here that “host” factors – genetic background, exposures and other modulators of host state during life – are the “black box” that has been understudied from a mechanistic standpoint.

Evidence that host genetic background impacts cognitive presentation in DLB, PD, and PDD comes from two major sources: clinicogenetic studies linking candidate genetic variants to cognitive trajectories in PD, and genome-wide association studies (GWAS) aimed at finding variants that associate with PDD or DLB. APOE genotype has been widely linked to cognition in LBD, with carriers of the APOE4 allele more likely to develop cognitive decline and dementia, in studies from our group and others.^179–182 Similarly, GBA mutations have been widely reported to associate with poorer cognition in PD patients.^183–187 Finally, since SNCA was first linked to familial PD, carriers of SNCA mutations (and multiplications) have been noted to have prominent cognitive symptoms,^28,188 with common variants at the SNCA locus also associating with cognitive outcomes in PD. More recently, GWAS have been performed to discover, in an unbiased manner, common genetic variants that associate with cognitive phenotype in PD. These GWAS have confirmed associations for GBA and APOE, and nominated other loci near RIMS2, TMEM108, and WWOX as novel variants associating with cognition in PD.¹⁸⁹ Additionally, GWAS performed in DLB have also confirmed associations between GBA, APOE, and SNCA and this cognitive-symptom-first form of LBD, while nominating loci near BIN1 and TMEM175 as DLB risk loci.¹⁹⁰

Notably, genotypes at the BIN1 locus, first associated with risk of AD by GWAS, have also been shown by our group to associate with the presence of AD neuropathological change (ADNC) at autopsy in LBD cases,¹⁹¹ again underlining the importance of AD pathology in LBD patients. Indeed, we have argued that the high prevalence of concomitant AD pathology in LBD patients at autopsy – with only 43/208 (20.7%) of LBD patients in our study demonstrating no ADNC at autopsy, and 79/208 (37.98%) demonstrating intermediate to high levels of ADNC¹⁹¹ – raises the question of whether PD patients might represent the ideal pre-AD group in which to target mechanisms leading to AD pathology therapeutically.⁶¹ A role for AD pathology in determining the cognitive course of PD is also supported by biofluid biomarker studies (beta-amyloid and tau biomarker levels associated with AD are enriched in PD patients with cognitive deficits)^192,193 and imaging studies (an AD-like structural MRI pattern is enriched in PD patients with cognitive deficits, who also have higher levels of beta-amyloid positivity by PET).^194–197

Host exposures have been linked to the development of PD for decades, with consistent epidemiological evidence for exposure to pesticides (particularly paraquat), the solvent trichloroethylene (TCE), and mild-to-moderate head injury contributing to increased risk for PD.^198,199 Moreover, animals exposed to paraquat,²⁰⁰ TCE.^201,202 or traumatic brain injury (TBI)²⁰³ have been reported to develop varying degrees of neurotoxicity or neurodegeneration. In addition, caffeine and tobacco use inversely associate with risk for PD in many epidemiological studies.²⁰⁴ Whether these host exposures also increase risk for cognitive decline in PD is less well-studied, but it is worth noting that traumatic brain injury (TBI) has also been associated with subsequent development of dementia.²⁰⁵ Mechanistically, the disruption of the blood brain barrier,²⁰⁶ resultant inflammatory cascade,²⁰⁷ and microglial activation²⁰⁸ that may occur with TBI may certainly modulate the aSyn spreading events discussed earlier in this review. In addition to these environmental exposures, alterations in host state such as altered glucose metabolism²⁰⁹ or inflammation²¹⁰ may impact both the development of LBD and cognitive decline in the LBD. Indeed, we recently found, in a screen of 940 plasma proteins, that modestly elevated C-reactive protein (CRP) levels, often indicative of mild inflammation,²¹¹ associated with slower rates of cognitive decline in PD.²¹² Whether glucose metabolism and systemic inflammatory state relate mechanistically to the aSyn aggregation and spreading events discussed in this review is not currently clear.

Until recently, most work on the impact of host genetics and exposures on cognitive decline in PD has been correlative. In the last few years, however, detailed mechanistic studies have begun to link specific host factors to cellular events in the pathogenesis of cognitive decline in PD. In 2020, two separate groups demonstrated in mouse models that APOE genetic background modulated development of aSyn pathology through pathways independent of amyloid-beta. Davis, Holtzman, and coauthors found that APOE4 mice demonstrated accelerated spread of aSyn pathology following intrastriatal injection of aSyn PFFs, compared to APOE2 or APOE3 mice.²¹³ Zhao, Bu, and coauthors delivered SNCA via adenop-associated virus, finding that APOE4 mice showed increased aSyn pathology and more motor and cognitive deficits than APOE2 or APOE3 mice.²¹⁴ These two studies provide a mechanistic counterpart for the observation that human APOE4 carriers with PD consistently show accelerated cognitive decline.^179–182 In an example involving a different host gene, Wie, Ren and colleagues reported that missense mutations in TMEM175, linked by GWAS to risk for PD and for DLB as noted above, impair the function of the encoded lysosomal potassium channel.²¹⁵ This organelle-level impairment in turn leads to vulnerability to damage in neurons, accelerated spreading of aSyn pathology in neurons and mice, and accelerated rates of cognitive and motor decline in PD patients.²¹⁶ Importantly, the TMEM175 lysosomal channel defect may be targetable, since treating cells with a downstream pathway activator restored channel function. In a final example already mentioned in this review, we have shown that PD risk genotypes at the GWAS-nominated GPNMB locus associate with higher levels of GPNMB expression. GPNMB expression, in turn, is both necessary and sufficient for cellular internalization of fibrillar aSyn, and subsequent development of aSyn pathology in cortical neurons.¹³⁴

These investigations in disease models amenable to manipulation, together with the clinicogenetic candidate gene studies and the GWAS cited above, provide strong support for the concept that host factors mechanistically impact the development of cognitive decline in LBD. Moreover, we underline three aspects of the evidence to date. First, multiple different approaches have converged on the same loci (GBA, APOE, SNCA, BIN1, TMEM175), increasing confidence that these host factors reliably influence risk for cognitive decline, across many cohorts and research strategies. Second, early mechanistic studies of APOE, TMEM175 and GPNMB demonstrate that biological pathways involving these genes may be amenable to therapeutic targeting, suggesting that investigating host genetics in disease-tailored, mechanistic experiments could help the field develop strategies for slowing or stopping cognitive decline in the LBD. Finally, for therapies aimed at GBA, APOE, TMEM175, GPNMB – or, for that matter, individuals who have been exposed to pesticides or TBI – a tailored approach matching the individual (who has a particular genetic or exposure profile) to the appropriate intervention is likely to be most effective.

Concluding Remarks

Cognitive impairment in PD is common, existing management options are limited, and the burden PD-related cognitive impairment and PDD impose on patients, caregivers, and health care systems are projected to grow at a breathtaking pace. As a result, we urgently need to understand underlying mechanisms and whether these mechanisms can be therapeutically targeted. Twenty years after its initial formulation, a refined version of the “Braak hypothesis” of pathological aSyn spread from brainstem to limbic and neocortical regions continues to inform. Specifically, evidence that pathological forms of aSyn may template naïve forms of aSyn in a prion-like manner, that pathological aSyn may exist as different “strains” with distinct conformations and behaviors, and that pathological aSyn may spread from cell-to-cell within the brain, eventually reaching the limbic and cortical regions responsible for many cognitive functions, is strong. At the same time, the frequency with which LBs are found in cognitively-unimpaired individuals, and the overall patterns of LB pathology in human autopsy studies, highlight areas that are poorly explained by an aSyn-focused, “Braak hypothesis” view. Specifically, we do not understand why some individuals with cortical aSyn pathology have PDD, while others do not show obvious cognitive symptoms. Similarly, we do not understand why some individuals with abundant brainstem aSyn pathology develop cortical aSyn pathology, while others do not (or why other individuals with abundant cortical aSyn pathology appear to have “bypassed” the brainstem entirely).

We argue that “host” factors – individual genetic characteristics and exposures or other modulators of host state during life – are the likely “black box” where answers to these questions may be found (Figure 3). We provide evidence from recent mechanistic work on APOE, TMEM175, and GPNMB, all genes in which common variants are known to associate with PD or with cognition in PD. Most encouraging, these recent developments suggest that pathways implicated by these “host” factors could be therapeutically exploited, offering more targets for intervention.

Figure 3. Factors influencing heterogeneity of cognitive outcomes in the Lewy body diseases.

Inherent pathologic features of “proteinopathy” (as exemplified by aSyn strain, or amount of co-existing AD pathology) are modulated by individual host features (genetic background and environmental exposures throughout life), resulting in varying cognitive presentations. PD-NC = Parkinson’s disease with normal cognition, PD-MCI = Parkinson’s disease with mild cognitive impairment, PDD = Parkinson’s disease with dementia, DLB = dementia with Lewy bodies.