Visual hallucinations in Parkinson’s spotlight on central cholinergic dysfunction

Authors: Anna Ignatavicius, Elie Matar, Simon J G Lewis

Abstract

Visual hallucinations are a common non-motor feature of Parkinson’s disease and have been associated with accelerated cognitive decline, increased mortality and early institutionalization. Despite their prevalence and negative impact on patient outcomes, the repertoire of treatments aimed at addressing this troubling symptom is limited. Over the past two decades, significant contributions have been made in uncovering the pathological and functional mechanisms of visual hallucinations, bringing us closer to the development of a comprehensive neurobiological framework.

Convergent evidence now suggests that degeneration within the central cholinergic system may play a significant role in the genesis and progression of visual hallucinations. Here, we outline how cholinergic dysfunction may serve as a potential unifying neurobiological substrate underlying the multifactorial and dynamic nature of visual hallucinations.

Drawing upon previous theoretical models, we explore the impact that alterations in cholinergic neurotransmission has on the core cognitive processes pertinent to abnormal perceptual experiences. We conclude by highlighting that a deeper understanding of cholinergic neurobiology and individual pathophysiology may help to improve established and emerging treatment strategies for the management of visual hallucinations and psychotic symptoms in Parkinson’s disease.

Introduction

Parkinson's disease (PD) is the second-most common and fastest growing neurodegenerative cause of disability worldwide.^1^ In addition to the cardinal motor symptoms which are primarily associated with dopaminergic degeneration in the substantia nigra,^2^ approximately 60% of patients will develop PD-associated psychosis within 12 years from diagnosis.^3^ Visual hallucinations (VH) are the most frequent feature of PD psychosis and have been associated with a more rapid conversion to dementia, increased mortality and are the strongest predictive factor for institutionalization.^4-6^ VH typically progress along a clinical spectrum from visual misperceptions to well-formed complex hallucinations and can evolve into a more pervasive psychosis with delusions, although this order is not fixed.^7^ Despite their prevalence and impact on patient outcomes, treatments aimed at addressing this troubling symptom are currently limited, representing an unmet clinical need.

Early investigations into the cause of VH in PD suggested they were a side-effect of levodopa therapy. However, hallucinatory phenomena are known to occur in medication-naïve patients^8^ and high-dose intravenous levodopa infusion has not been shown to induce VH in patients who regularly experience this symptom.^9^ Thus, whilst dopaminergic treatments may exacerbate VH in some patients, it is unlikely to be the primary driving factor contributing to their complex aetiology. In addition to dopaminergic deficits, PD patients also exhibit abnormalities in other neuromodulatory systems,^10^ and there is now considerable evidence that degeneration within the cholinergic system in particular may be linked to the emergence of VH. This is supported by the observations that anticholinergics can induce VH not just in PD but also in other clinical populations and healthy adults.^11,12^ Furthermore, cholinesterase inhibitors (ChEIs), which increase cholinergic levels in the synapse, can reduce psychotic symptoms in PD.^13^

In this review, we focus on the contribution of cholinergic dysfunction to the pathophysiological processes underpinning VH in PD. First, an overview of the central cholinergic system is provided, followed by a comprehensive summary of the key findings from post-mortem and in vivo research, examining the micro- and macroscopic alterations observed in PD patients with VH. We then evaluate how a disrupted cholinergic system may be the common denominator that reconciles some of the overlapping and disparate mechanisms outlined by the major theoretical models that have been proposed to explain VH. We conclude by describing how a greater appreciation of cholinergic pathophysiology may facilitate the development of novel treatment strategies.

Neuroanatomy and function of the central cholinergic system

Major cholinergic nuclei and their projection targets

Two major types of cholinergic neurons exist within the CNS: (i) thinly- or un-myelinated projection neurons, which connect different cortical and subcortical brain regions; and (ii) tonically active interneurons, which provide dense, localized innervation.^14^ Cholinergic projection neurons originate primarily in the basal forebrain and brainstem structures where they are co-distributed with populations of glutaminergic and GABAergic neurons. Cell clusters within these regions exhibit a topographical arrangement with extensive projections targeting selective brain regions (Table 1).

The magnocellular basal forebrain complex provides the majority of cholinergic input to the neocortex and limbic system.^22^ The largest of the basal forebrain nuclei is the nucleus basalis of Meynert (NBM), which lies within the substantia innominata and is composed of approximately 90% cholinergic neurons.^15,16^ Projections from the NBM have been associated with the regulation of a wide range of important cognitive processes including attention, memory formation, perceptio and arousal.^22,23^ Despite its irregular shape and lack of distinct anatomical boundaries, neurons within the NBM show segregation across rostrocaudal, mediolateral and dorsoventral axes with their projections arranged in a diffuse, yet organized manner.^15,16,24^ Therefore, the NBM can be further parcellated into functional subdivisions based on anatomo-topographic characteristics.^16,22^ The anterior NBM projects to the frontal cortex, cingulate and amygdala, the intermediate NBM to the parietal and occipital cortices and the posterior NBM to the temporal lobe.^22^

Within the brainstem, the pedunculopontine nucleus (PPN) contains the highest proportion of cholinergic neurons in this region, with reciprocal connections to the basal ganglia and most, if not all, of the thalamic nuclei.^18^ It also projects to the NBM and receives afferent input from the cerebral cortex, the locus coeruleus and the dorsal raphe.^25^ The PPN plays an important role in regulating sleep, arousal, as well as posture and locomotion.^18,26^ Additionally, it is involved in visual processing and attention, both via direct innervation of the forebrain cholinergic system and indirectly through activation of thalamocortical projections.^27,28^

The spatiotemporal dynamics of cholinergic signalling

The spatiotemporal resolution of cholinergic signalling is defined by the distribution and type of cholinergic receptors in the brain. Nicotinic receptors (nAChR) are predominantly associated with rapid depolarization, while the slower effects of acetylcholine are mediated by the longer latency of metabotropic muscarinic receptor (mAChR) activation.^14^ Response dynamics also vary based on receptor subtype and the downstream signalling cascades triggered by their activation, refining the specificity of cholinergic neurotransmission.^23^

Nicotinic receptors are ligand-gated ionophores with a pentameric structure composed of α2–α10 and β2–β4 subunits.^29^ The two major subtypes expressed in the CNS are homo-oligomeric α7 nAChRs and hetero-oligomeric α4β2 nAChRs, each exhibiting distinct pharmacological profiles and kinetics.^30^ For instance, α7 nAChRs are capable of eliciting both fast excitatory responses and sustained cellular effects due to their significant calcium permeability.^31^ Among the five mAChR subtypes identified in the mammalian brain, M1, M3, M5 are preferentially coupled to Gq protein subunits and are associated with increased neuronal excitability.^32^ Conversely, M2 and M4 are coupled with Gi/o protein subunits which modulate ion channel functions through a relatively faster membrane-delimited pathway.^32,33^ Presynaptically, M2/M4 mAChRs at cholinergic terminals act as inhibitory auto-receptors, thus play a role in regulating cholinergic activity.^14^ Cholinergic signalling is further controlled by receptor affinity for acetylcholine, as well as the rate of localized clearance determined by the extracellular concentration of acetylcholinesterase (AChE).

Despite evidence highlighting acetylcholine’s diverse actions on target neurons, controversy still surrounds whether cholinergic signalling acts via wired and phasic synaptic transmission or in diffuse and tonic mode via volume transmission (see Disney and Higley^34^ and Sarter and Lustig^35^). However, sophisticated amperometry and mesoscopic imaging studies have revealed that acetylcholine release has coordinated phasic and tonic components with specific effects on behavioural outcomes, highlighting the complexity of its actions at both a local and global level.^36-38^ Therefore, cholinergic signalling is capable of facilitating low-resolution phenomena such as cortical arousal but also processes defined by spatiotemporal precision, such as cue detection, sensory gating and memory encoding.^36,37,39^

Neuropathological and neurophysiological correlates of visual hallucinations in Parkinson’s disease

Post-mortem neuropathology

The pathological hallmark of PD is the accumulation of Lewy bodies (LB) and Lewy neurites, which are abnormal eosinophilic protein aggregates found in neuronal perikarya and neuronal processes of selective brain regions.^40^ These inclusions are primarily composed of misfolded α-synuclein and were first identified in the NBM of post-mortem PD brain tissue by Friedrich Lewy in 1913.^22^

More contemporary analyses have confirmed the presence of LB in the NBM accompanied by cholinergic cell loss in patients with PD dementia (PDD).^22,41,42^ Although LB are also found in the NBM of PD patients without dementia, concomitant degeneration is not always evident.^41^ Nevertheless, reduced choline acetyltransferase activity has been observed in the frontal cortex in non-demented PD post-mortem tissue even in the absence of considerable neuronal loss.^41^ The exact mechanism by which LB pathology may lead to cholinergic dysfunction remains uncertain, but it has been hypothesized that intracellular α-synuclein deposition disrupts neurotransmitter production by sequestering rate-limiting enzymes necessary for acetylcholine synthesis.^43^ Cholinergic deficiencies may therefore develop insidiously before the onset of obvious neuronal necrosis.

In addition to the NBM, LB pathology has also been detected in the PPN of PD patients.^44^ Notably, significant atrophy of the PPN has been reported following post-mortem histopathological assessment of PD patients who experienced VH.^45^ However, in this particular study, a comparison group non-hallucinators was not included, making it difficult to determine whether PPN pathology directly contributes to VH.

Several studies have established a clinicopathological correlation between VH and the presence of LB within the limbic regions including the amygdala, parahippocampal cortex and inferior temporal cortex.^41,42,46-48^ Within the amygdala, LB deposition is highly selective and has been found in the cortical and basolateral nuclei in non-demented PD brains.^46^ Although no substantial loss in volume was observed, the proportion of LB-containing neurons in the basolateral amygdala was nearly double in patients with VH.^46^ The amygdala has been associated with enhancing visual awareness, as well as redirecting attention to stimuli with motivational relevance, a process mediated by cholinergic projections from the NBM.^49,50^ Consequently, increased LB density in both of these regions could result in impaired visual attention and potentially contribute to the mechanisms underscoring VH.

Alpha-synuclein positive inclusions have also been identified in the intralaminar thalamic nuclei in both PD and PDD.^51^ Greater atrophy of the intralaminar nucleus was associated with the presence of VH, but not dementia, in both patient groups. Degeneration within the intralaminar nucleus and loss of cholinergic innervation from the PPN could potentially disrupt cortico-cortical information transmission due to connections with the amygdala and frontoparietal cortical regions.^52^

An important caveat to note is that aberrant α-synuclein aggregation, per se, may not be the only neuropathological substrate of VH in PD as they have been found to correlate to changes in areas with relatively mild LB burden.^53^ For example, atrophy of the thalamic mediodorsal nucleus has been associated with VH in both PDD and dementia with Lewy bodies (DLB) despite minimal α-synuclein immunoreactivity.^53^ Differences in neuronal number or size did not significantly contribute to the observed volume loss, suggesting that pathological changes may be occurring at the level of the synapse or neuropil. Decreased choline acetyltransferase activity has been correlated with a reduction in synaptic markers in the primary visual cortex in DLB, another region where the presence of LB is relatively low.^47,54,55^ Speculatively, degeneration within cholinergic networks may be partially responsible for downstream synaptic changes in areas without overt neuropathology.

The co-occurrence of amyloid-beta (Aβ) plaques and neurofibrillary tangles alongside LB pathology has also been documented in PD patients with VH, particularly in the frontal, parietal and hippocampal regions.^56^ This is consistent with the observation that reduced baseline CSF levels of Aβ42 is a risk factor for early-onset PD-psychosis, suggesting that VH may predict the precipitation of Alzheimer’s disease (AD)-type pathology in PD patients.^56,57^ Moreover, higher cortical Aβ burden has been associated with a shorter latency to dementia in PD and is positively correlated with LB pathology, indicating a potential shared pathological mechanism linking VH and cognitive decline.^58,59^ Evidence from animal models of AD suggests that a dysfunctional cholinergic system could be associated with increased cortical amyloid deposition. Lesions to the NBM and reduced cortical ACh release have been shown to elevate amyloid precursor protein synthesis in the cortex, and the loss of M1 mAChRs can increase the level of amyloidogenic Aβ peptides.^60,61^ Therefore, greater coincident Aβ pathology may represent a cholinergic phenotype of PD associated with a more severe cognitive and neuropsychiatric symptomatic profile.^62^

While histopathological studies provide valuable correlational evidence relating to the distribution of pathology and VH, they are often limited by the lack of standardized antemortem assessment and inevitably reflect changes associated with end-stage disease. Therefore, viewing these findings within the context of structural, functional and metabolic alterations observed in vivo provides additional insights into the neuropathological changes unique to VH.

Molecular imaging

Acetylcholinesterase and vesicular acetylcholine transporter (VAChT) PET imaging studies allow for in vivo assessment of the integrity and density of regional cholinergic synapses.^62,63^ Reduction in cortical and thalamic uptake is indicative of degeneration in the ascending cholinergic systems from the NBM and PPN, respectively.^64^ A recent AChE PET study in PD patients found a distinct association with NBM volume and cortical cholinergic activity in the parietal, occipital and temporal regions, as well as the anterior cingulate cortex (ACC).^65^ Similarly, an earlier investigation also reported reduced cortical AChE activity in patients with PD, particularly within the temporal and occipital cortices.^66^ This reduction was more significant in subjects with VH compared to those without.^66^

In PD, regional brain changes in VAChT activity have also been identified in the occipital, parietal and posterior temporal regions, as well as regions of the major attentional network hubs including the cingulate cortex, bilateral insula, operculum and visual thalamus, underscoring the vulnerability of both the NBM and PPN cholinergic systems.^67,68^ A loss of cholinergic innervation to the visual thalamus, including the lateral geniculate nucleus (LGN), may impair attentional processing of visual information.^69^ In fact, both reduced global cortical VAChT and AChE activity have been correlated with impaired performance in attention, executive and visuospatial domains,^62^ all of which have been previously associated with VH in PD.^70,71^

Cholinergic denervation is believed to follow a posterior to anterior progression as neurons with poorly myelinated, distally projecting axons are more vulnerable to pathology and oxidative stress. Consequently, terminals at the posterior cortices would be most susceptible to a loss of cholinergic input.^72,73^ This hypothesis has been recently contested in a longitudinal study that examined the topographical cortical cholinergic changes in PD patients over a 4–8-year period.^64^ The findings revealed a pattern of progressive denervation in posterior brain regions following a caudo-rostral gradient, with relative preservation of early visual areas. The decline of AChE activity in the posterior cortices largely corresponded with reductions in regional cerebral blood flow (CBF). This indicates that posterior cholinergic dysfunction may be the result of more diffuse cortical synaptic loss, as regional CBF strongly correlates with regional synaptic density.^74,75^ Conversely, in the mid-frontal cortical regions, cholinergic denervation showed an opposite pattern. Here, reductions in AChE activity progressed along a rostro-caudal gradient, occurring disproportionately to changes in regional CBF. Therefore, anterior cholinergic deficits may reflect the preferential loss or dysfunction of cholinergic synapses. These results highlight that the pathological basis of cholinergic denervation likely involves differential, region-dependent processes, including the selective degeneration of cholinergic basal forebrain neurons and more general neurodegenerative changes occurring at the cortical level. The pattern of cholinergic denervation included regions within the visual processing streams and critical nodes within the default mode network (DMN), as well as the both the ventral and dorsal attentional networks.

The DMN is associated with mind-wandering, introspection and episodic memory retrieval,^76^ while the dorsal attention network (DAN) is involved in voluntary, top-down attentional control.^77^ The ventral attention network (VAN) rapidly engages attention to novel salient stimuli and facilitates the transition from internal to external attentional processing.^78^ The progression of cholinergic deficits in these network regions, along with altered activity in the ventral and dorsal visual streams, might impair the ability to effectively perceive and interpret visual stimuli.^64^ Indeed, each of these networks has been implicated in existing mechanistic models of VH, as discussed further below.

Nicotinic and muscarinic receptor molecular imaging in patients with PD has revealed significant alterations in regional receptor binding in both cortical and subcortical areas.^79^ Several single-photon emission computerized tomography (SPECT) studies have found widespread decreases in binding of ligands targeting the β2-containing nAChR in the thalamus, parietal and temporal cortices, as well as the amygdala and hippocampus.^80-83^

Bidirectional changes in regional mAChR expression have been observed in PD patients, which might reflect compensatory responses in response to cholinergic deficits in other regions.^84,85^ Colloby et al.^84^ used a SPECT ligand selective for M1/M4 mAChRs to examine alterations in receptor expression within interconnected brain regions. Results showed a pattern of decreased M1/M4 binding within the ACC, lateral temporal cortex, striatum, parahippocampal gyrus and visual association areas with concomitant preserved/increased binding occipitoparietal cortex, as well as regions within the DMN. The distribution of altered muscarinic receptor expression partially overlaps with findings from an earlier investigation in patients with PDD.^86^ However, in this cohort, decreased M1/M4 receptor binding extended to include the basal forebrain, medial temporal cortex, as well as the insula, a key hub within the VAN. Thus, the precipitation of VH and conversion to dementia may be associated with alterations to the cholinergic networks that subserve visual attention, perception and memory processes.

Structural MRI

Converging evidence from neuroimaging studies suggests that the emergence of VH in PD cannot be ascribed to a localized lesion within the brain but is rather associated with dysfunction of multiple brain areas.^87,88^ Despite significant heterogeneity across structural imaging studies, some consistent grey matter changes associated with VH in PD have been identified.^89^ These regions include the primary visual, extra-striatal and visual parietal cortical regions, as well as subcortical structures such as the thalamus and hippocampus.^87,89-93^ Reduced grey matter volume has also been observed in the PPN^94^ and the substantia innominata^95^ in PD patients with VH compared to those without. In a recent longitudinal study, it was found that lower baseline NBM volume in de novo PD patients was associated with the subsequent development of VH.^96^ This reinforces the cholinergic contribution to VH and suggests that the pathogenic processes and vulnerabilities affecting cholinergic transmission may occur even in the early stages of disease.

Meta-analytical approaches have helped to identify structural changes in brain regions associated with VH that might go undetected in smaller cross-sectional studies. A recent meta-analysis pooled subject-level structural imaging data from 493 PD patients (135 with VH) acquired by multiple research groups while accounting for possible confounding factors due to different scanning locations.^88^ Widespread alterations in brain structure were associated with VH involving regions within the frontoparietal, occipital and temporal lobes. Structural covariance of differences in cortical thickness and surface area between hallucinators and non-hallucinators revealed that interregional correlations overlapped with functional networks comprising the DAN and VAN. Conversely, no difference in surface area was found in key DMN regions suggesting that the DMN may be relatively preserved compared to the DAN and VAN in patients with VH. A subset analysis also revealed significant volume reductions in hippocampus and amygdala, which were associated with VH independent of cognition, disease duration and severity. Another recent meta-analysis that used coordinate-based network mapping found that while the areas of grey matter atrophy associated with PD-hallucinations were diverse and inconsistent across individual studies, these abnormalities localized to a common network centred around the LGN.^97^ This network was specific to VH, distinct from the identified networks that correlated with PDD and PD-associated mild cognitive impairment (MCI).

Although studies investigating white matter alterations in PD-hallucinators are limited, there is evidence that dystrophic axonal changes may occur in PD before significant grey matter deterioration. A recent diffusion-tensor imaging study found reduced afferent white matter tract integrity from the NBM to the parietal and occipital cortices in PD patients with VH compared to those without, despite similar normalized NBM volumes. This suggests a potential involvement of the lateral NBM pathway.^98^ Previous histological and tractography studies have demonstrated that the capsular division of the lateral tract traverses the external capsule to the parietal and temporal cortices via the posterior thalamic radiation and superior longitudinal fasciculus, and through the splenium of the corpus callosum to merge with the occipital cortex.^24,99^ Therefore, cholinergic neurotransmission may be vulnerable to insults to white matter pathways, even without substantial neuronal loss within the NBM.^24^ Results from a longitudinal study that investigated whole-brain and thalamic structural changes found significant white matter changes within the corpus callosum and posterior thalamic radiation in PD patients with VH at baseline.^100^ Compared to patients without VH, PD-hallucinators also showed reduced white matter tract integrity from the right mediodorsal thalamus at baseline, with no significant differences in cortical thickness. However, at the 18-month follow-up, multiple posterior and thalamocortical white matter tracts were affected, as well as atrophy of the right mediodorsal thalamic nucleus and regional reductions in cortical thickness. Together, these findings emphasize early posterior tract degeneration as well as changes to both cholinergic and thalamic projections in the development of VH. The loss of structural connectivity seen in PD patients with VH might lead to maladaptive alterations within large scale functional networks, causing disrupted or abnormal information flow across attentional and perceptual systems.^101^

The selective vulnerability of specific neurotransmitter systems may play a role in the multifaceted pathological processes that contribute to regional atrophy and structural brain reorganization in PD patients with VH.^88,102^ Patient-specific modelling has revealed that interindividual variability in symptom severity within the visuospatial and memory domains is associated with cholinergic receptor-mediated biological mechanisms in PD.^102^ The neuropathological changes underlying these symptoms are influenced by muscarinic and nicotinic receptors, contributing to degenerative alterations associated with dendritic density, grey matter reductio and microstructural white matter integrity.

Functional MRI

Resting-state functional MRI (rs-fMRI) investigations comparing PD patients with and without hallucinations consistently indicate heightened DMN activity and abnormal coupling with the visual network.^91,103,104^ Notably, PD patients with mild hallucinations exhibit increased within-network resting-state functional connectivity between hubs of the DMN, as well as stronger connectivity between the posterior cingulate cortex and ventral visual processing areas, suggesting similar functional correlates may underscore both mild and well-structured hallucinatory phenomena.^104^ Increased resting-state functional connectivity has also been observed between the hippocampus and other regions of the DMN including the precuneus, as well as decreased functional connectivity between the hippocampus and regions within the dorsal and ventral visual pathways in PD patients with VH compared to those without. Spatial and non-spatial information from the dorsal and ventral streams converge on the hippocampus, where it plays a crucial role in complex visual processing necessary for contextual scene representation and visuospatial attention.^105-107^ These findings suggest that aberrant connectivity between the DMN and visual associative areas may contribute to the manifestation of VH.

Diffusion-map embedding is a non-linear dimension reduction method that maps resting-state global connectivity as a distribution of cortical nodes representing brain organization as a functional gradient.^108^ Nodes with similar functional connectivity are grouped closer together, whereas nodes that have dissimilar functional connectivity are grouped further apart.^108,109^ The compression of the unimodal (bottom-up, sensory regions) and heteromodal (top-down, higher order regions) gradient has been observed in PD patients with visual dysfunction^110^ and recently in patients with VH.^109^ Alterations in the functional hierarchy may reflect abnormal interactions whereby higher order cognitive regions influence early perceptual processes, increasing the vulnerability to hallucinate.

While rs-fMRI analyses can provide valuable insight about how functional brain networks communicate, the directionality of these interactions cannot be inferred. Dynamic causal modelling (DCM) uses a Bayesian framework to extrapolate information on the directional (causal) connection strengths between brain regions, referred to as effective connectivity.^111^ A modification well-suited to rs-fMRI data is spectral DCM, which parametrizes the spontaneous fluctuations in blood oxygen level-dependent signals, modelling dynamics between brain regions in the frequency domain rather than the time domain.^111,112^ Thomas et al.^112^ applied spectral DCM to rs-fMRI data in PD patients with and without VH and found that hallucinators show decreased bottom-up effective connectivity from the LGN to the primary visual cortex and increased top-down effective connectivity from the left prefrontal cortex (PFC) to the primary visual cortex and medial thalamus. The alterations in connectivity patterns were also associated with the severity of hallucinations, suggesting that both bottom-up perceptual deficits and overly influential top-down processing may together contribute to the erroneous interpretation of visual input. Evidence has shown that acetylcholine modulates thalamocortical excitability and cortico-cortical transmission during uncertainty, favouring stimulus-driven, bottom-up processing over context-bound, top-down processing.^14,113^ This is supported by findings that preserved thalamic cholinergic integrity in PD predicts signal detection under perceptual noise.^114^ Therefore, these functional changes within the perceptual hierarchy in PD patients with VH may be partially associated with dysfunctional cholinergic signalling.

Dynamic rs-fMRI techniques explore the degree of integration and segregation across the whole-brain network, and the transition between functional states has been shown to be influenced by distinct neuromodulatory neurotransmitters.^115,116^ A study that combined dynamic functional connectivity and network control analysis revealed that PD patients with VH spent more time in a segregated functional state and had fewer transitions between states in comparison to those without hallucinations.^117^ As temporal dynamics are constrained by the structural connectome,^118,119^ the predisposition to remain in a segregated state may be due to the reduced microstructural white matter integrity that has been previously observed in patients with VH.^100,101,120^ Furthermore, the ‘energy cost’ to transition between states was correlated with the expression of cholinergic (muscarinic and nicotinic), as well as serotonergic, GABAergic and noradrenergic receptors.^117^ This suggests that changes within the cholinergic system may play a role in altered state transitions in PD-hallucinators. Consequently, therapeutic agents that directly target cortical cholinergic receptors may help to correct the network imbalance seen in these patients.

Collectively, these findings are informative about the trait-level features that might predispose an individual to hallucinate. However, it is unclear if these network-level abnormalities also underlie the hallucinatory-state itself. Novel behavioural tasks, such as the Bistable Percept Paradigm (BPP), have been shown to be a viable surrogate of hallucinatory phenomena.^121-123^ The BPP involves processing a series of monochromatic images that represent either ‘bistable’ (ambiguous, i.e. amenable to more than one interpretation) or ‘stable’ (unambiguous, i.e. single interpretation) visual stimuli.^121^ PD patients with VH more frequently misperceive visual stimuli compared to their non-hallucinating counterparts, indicating additional features within stable images.^121^

Previous event-related fMRI studies using the BPP have highlighted the role of altered connectivity between attentional networks.^123^ Shine et al.^123^ found that PD-hallucinators showed greater activity within the VAN and DMN during misperceptions accompanied by increased functional coupling of the DMN and visual network. These patients also exhibited a relative inability to recruit the DAN when interpreting visual stimuli and the degree of decreased activation of the frontal and parietal hubs within this network was predictive of the strength of DMN-visual network coupling. These findings suggest that VH may reflect a transient state where a failure to engage exogenous attentional mechanisms results in an over-reliance on endogenous networks for the interpretation of visual stimuli.^124^

Perfusion and metabolic imaging

Metabolic fluorodeoxyglucose (FDG)-PET imaging studies comparing PD patients with and without VH have found decreased cerebral metabolism in the occipital, temporal and parietal regions in PD patients with VH.^125-127^ However, the evidence for metabolic alterations in the frontal regions has been less consistent. For example, some studies have reported reduced regional cerebral metabolic rate for glucose consumption (CMRglc) in scattered frontal areas together with posterior hypometabolism,^126^ while others have observed increased cerebral metabolism in the superior frontal gyrus in PD patients with VH, compared to those without VH.^125^ SPECT studies have revealed similar patterns of posterior hypoperfusion, as well as inconsistent evidence of regional hyperperfusion. Oishi et al.^128^ found regional CBF was reduced in the right fusiform gyrus and increased in the right superior and middle temporal gyri in PD patients with VH. Conversely, Matsui et al.^129^ observed decreased perfusion in the bilateral inferior parietal lobule, inferior temporal gyrus, precuneus and occipital cortex in PD patients with VH, relative to those without VH, with no regions displaying significant hyperperfusion.

The inconsistencies relating to regional CBF and CMRglc may be related to the fact that these nuclear imaging studies relied upon global mean normalization, which is only valid when there is no difference in global mean values between groups.^130,131^ This may have led to the appearance of artefactual increases in regions where absolute perfusion was, in fact, preserved, whilst simultaneously failing to detect lower magnitude decreases in CBF and CMRglc, particularly within subcortical structures.^131^

The pattern of perfusion abnormalities seen in PD patients with VH is likely the result of complex interactions among neuronal degeneration, functional disruptions, and alterations in neurotransmitter systems. Endothelial cells that control vascular tone and CBF receive innervation from cholinergic neurons that originate in the basal forebrain.^132^ Acetylcholine contributes to localized arterial vasodilation during neural activity via the activation of endothelial nitric oxide synthase stimulating the production of nitric oxide.^133,134^ Previous research has associated atrophy in the basal forebrain with widespread cortical hypometabolism in cases of MCI.^135^ This suggests that cholinergic degeneration may potentially impair neurovascular coupling mechanisms, and thereby brain function, in a manner independent from its effects on neuronal activity.^133^

Electrophysiology

Short-latency afferent inhibition (SAI) is an electrophysiological technique that serves as a proxy measure of cholinergic excitability by assessing an inhibitory circuit in the motor cortex dependent on central cholinergic activity.^136^ The SAI is evoked when transcranial magnetic stimulation is delivered to the contralateral motor cortex between delay intervals of 2–8 ms following peripheral nerve stimulation. Manganelli et al.^137^ found that majority of patients with VH exhibited SAI values that deviated from the normal range, whereas non-hallucinating patients displayed no SAI abnormalities. It is possible that the normal SAI in some PD patients may be related to the compensatory upregulation of mAChRs observed in molecular imaging studies,^84,85^ secondary to the loss of ascending cholinergic input to the frontal cortex from the NBM.

Both mild and well-structured VH in PD have also been associated with slowing of resting-state oscillatory brain activity, characterized by an augmentation in the theta (4–7 Hz) activity, alongside a reduction in gamma (>32 Hz) activity.^138,139^ In a source-based MEG study, PD patients with VH had significantly higher relative power in the theta band in all but the frontal brain regions, as well as decreases in both beta (16–31 Hz) power in the right temporoparietal region and gamma power in the bilateral frontal and limbic regions compared to patients without VH.^138^ These distinct spectral differences may be related to a greater cholinergic deficit in patients who experience VH. Administration of the cholinergic antagonist, scopolamine, has been shown to alter the spatial and temporal dynamics of brain oscillatory activity, resulting in a relative increase in theta power and decrease in alpha power in posterior brain regions.^140^ Taken together, the findings from these electrophysiological investigations further support central cholinergic dysfunction as a significant contributing factor to the pathological mechanisms that underpin VH in PD.

Cholinergic contributions to existing theoretical models of visual hallucinations in Parkinson’s disease

Over the past two decades, several hypothetical models have been developed in an attempt to understand the complex phenomenon of VH (summarized in Table 2).^141,142^ The functional mechanisms of these models have largely been shaped by the clinical observations at the time of their development (for comprehensive reviews, see Muller et al.^141^ and Collerton et al.^142^). Although now that there is a wealth of experimental data, there has been a struggle to develop a unified model that accounts for the multifactorial pathological and cognitive changes associated with the phenomenon of VH in PD. One of the challenges is to distinguish the integral components that are unique to VH and separate from symptomatic associations that may be related to shared or concurrent pathology occurring at similar disease stages. To overcome this, recent collaborative efforts have produced a harmonized consensus framework delineating a set of fundamental cognitive systems pertinent to VH, namely arousal, visual processing, memory and attention (Fig. 1A).^142^ In the following sections, we use this framework as a bridge to understand the mechanisms by which the alterations in the cholinergic system may produce VH. The role of the cholinergic system is mapped to each relevant cognitive process (Table 3 and Fig. 1B), thereby offering cholinergic dysfunction as a core and unifying neurobiological substrate of the interactions contained in the functional framework.

Arousal

Arousal is a multifaceted concept that represents a spectrum of functional brain states associated with levels of consciousness and the sleep-wake cycle. For adequate cognitive control and volitional behaviour, arousal levels require regulation and adaptation depending on environmental demands.^151^ The role of the cholinergic system in regulating arousal is well-established and is mediated by the modulatory influence of acetylcholine on cortical activity.^39,152^ Evidence highlights the involvement of mAChRs in global arousal regulation, which may help to augment the effects of rapid and phasic cholinergic signalling associated with appropriate cue detection and attention.^153,154^ Therefore, tonic cholinergic activity enhances cortical arousal in order to bias processing towards goal-relevant or perceptually salient stimuli, affecting intentional expectancies and prediction errors.^36,39,155^ Inadequate arousal signals to the thalamus, and cortex, could potentially impair the appropriate integration of bottom-up and top-down inputs and allow for intrusions of internal imagery, particularly when the quality of sensory data is diminished.

Heightened activation of cholinergic neurons in the basal forebrain and PPN is observed during wakefulness and REM sleep^156-158^ and atrophy within these regions has been linked to sleep disturbances,^159,160^ as well as VH.^45,94,95^ Furthermore, SAI is reduced in both PD patients with VH,^137^ and PD patients with RBD,^161^ suggesting a potential shared cholinergic basis. Consequently, a dysfunctional cholinergic system may serve as the gateway through which dream imagery infiltrates conscious perception. In this setting, aberrant signalling between pathologically affected components of the sleep-wake circuitry including the cholinergic REM promoting nuclei may trigger abnormal transitions between wakefulness and REM sleep, resulting in the intrusion of oneiric imagery into consciousness. This may explain work highlighting the relationship between VH, sleep disorders and sleep-wake transitions.^143,162-164^

If VH do arise from dream intrusions associated with sleep-wake transitions, then this is unlikely to be the only mechanism. Despite some evidence to suggest that hallucinations and REM sleep are temporally related,^143,165^ polysomnography studies have shown that hallucinatory experiences can also occur in an unequivocal wakeful state.^166^ Moreover, dream imagery is phenomenologically different to the repetitive and stereotypical nature of VH in PD which is often experienced as a neutral single object, persons or animals that tend to occupy the centre of the visual field.^167,168^ Alternatively, it has been hypothesized that there is more than one type of hallucinatory experience.^146,169^ Hallucinations that occur on the edges of sleep may be associated with dream intrusions and therefore more likely to be associated with sleep disturbances.^169^ However, VH that occur during wakefulness might involve more multifactorial influences on perception. These pathological mechanisms could include fluctuating vigilance and impaired attentional control caused by degeneration of the brainstem and basal forebrain cholinergic systems which when combined with visual processing deficits, leads to complex but stereotyped VH.^146,170^

Visual processing

The visual system receives cholinergic innervation from both cortical projections from the NBM and thalamic projections from the PPN.^15,16,171^ This dual-source modulation has implications for how visual information is subsequently processed along the perceptual hierarchy.^171^ Normally, cholinergic activity is thought to increase neuronal signal to noise ratio (SNR) within cortical networks, making neurons more responsive to external stimuli.^14,172,173^ Within the visual cortex, it is proposed that cholinergic innervation arising from the NBM desynchronises local field potentials, leading to an increase in higher frequency oscillations and a decrease in lower frequency oscillations.^174^ This desynchronization is synonymous with a high conductance state, which reduces noise from spontaneous neuronal fluctuations, resulting in increased signal amplitude and enhancing the detection of sensory signals.^175,176^ Activation of nAChRs and mAChRs on interneurons in the visual cortex ultimately leads to an increase in the SNR by suppressing intracortical communication.^14,177^ A loss of cholinergic activity to the occipital regions^64^ may therefore impair visual processing leading to an overreliance on memory contents or internally generated imagery to ‘fill in the gaps’ where it is then interpreted as veridical visual perception.

Reduced cholinergic input into the thalamocortical circuitry may also simultaneously degrade the reliability of sensory input. Cholinergic projections to the thalamus arising from the PPN act to enhance the precision of sensory signals by regulating thalamocortical interactions. This process is potentially mediated via M2 mAChRs and α7 nAChRs as part of the thalamic reticular nucleus (TRN),^178-180^ which modulates information flow in order to increase the salience of relevant stimuli and inhibit less salient stimuli or sensory noise. The TRN is a sheath of GABAergic inhibitory neurons around the thalamus that receives dense innervation from the PPN.^181^ It is thought to play a key role in the gating of visual stimuli due to its capacity to inhibit relay neurons of the LGN, therefore determining to which extent visual information from the retina is propagated to the visual cortex for further processing.^171,182,183^ VH in PD patients have been associated with atrophy of the PPN and its thalamic projections,^171^ suggesting that a lack of cholinergic innervation to key thalamic nuclei may account for a failure to update prior beliefs when sensory evidence is inconsistent with visual percepts.

Memory

Episodic memory deficits are frequently observed in individuals with PD who experience VH.^168^ The Reality Monitoring model emphasizes that accurate recollection of specific details from the encoding event is crucial for appropriate source attributions relying solely on familiarity is not a reliable means to determine the origin of information.^144^ Therefore, the risk of misperceiving internally generated events as veridical perception increases if the recollection of an encoding event is impaired.^144^

Associative recognition memory relies on the integration of information between the medial PFC and the hippocampus and parahippocampal regions.^184,185^ Rodent models have elucidated that nAChRs might play subtype specific roles in associative recognition memory. Selective receptor antagonism revealed that α7 nAChRs are associated with the encoding of events and the induction of long-term potentiation of hippocampal-prefrontal synapses, while the α4β2 nAChRs are crucial for the retrieval of associative memory.^186^ On the other hand, mAChRs have been shown to operate synergistically to facilitate both the encoding and retrieval processes.^187,188^ Cholinergic denervation has been observed within the temporal regions in patients with PD,^64^ with decreased M1/M4 binding detected within the parahippocampal gyrus,^84^ alongside reduced nAChR density in the hippocampus.^83,189^

Cholinergic signalling has a bidirectional influence on the neural dynamics within the cortical microcircuits that support memory encoding and retrieval.^190,191^ High cholinergic levels in the hippocampus promotes theta oscillations which bias the system towards external sensory processing and facilitates the updating of memory with novel information about the environment.^192,193^ Conversely, low cholinergic tone facilitates the generation of sharp wave ripples which support the recall of autobiographical and semantic memory and drives hippocampal-cortical interactions.^194,195^ In the context of visual perception, the hippocampus supplies memory-based expectations to the visual cortex and is essential for the application of perceptual priors.^196^ Therefore, alterations in cholinergic input to the hippocampus could tip the balance in favour of internally driven processes resulting in an over-processing of top-down generated information over bottom-up sensory stimuli.

Although memory deficits can be seen in the early stages of PD, taken alone, they are not sufficient to explain the emergence of VH. Instead, VH likely results from an interaction between dysfunctional visual and attentional processing, along with a reduced capacity to bind contextual components of memory together.^197^ In fact, recent evidence suggests that hippocampal networks may be more relevant to the presence of dementia than to VH.^97^ This does not preclude the involvement of the hippocampus, as both functional and structural alterations have been observed in patients with VH, even after controlling for cognition.^88,90,91^ Therefore, the hippocampus might contribute to the mechanisms of VH independently from its role in cognitive decline. Pathologically, PDD has been associated with more severe hippocampal LB burden^41,42^ and diffuse atrophy affecting the entire structure,^90^ which may explain the greater memory impairment seen in these patients. In contrast, VH without concomitant dementia has been linked to more discrete changes within specific hippocampal subfields.^90,91,198^ These alterations could reflect a selective loss of cholinergic innervation and subsequent synaptic degeneration rather than neuronal damage from localized intracellular pathology.

The functional connectivity of the DMN with hubs within the medial temporal lobe highlights its significant involvement in memory processes. Several theoretical models propose that overactivity of the DMN is related to the emergence of a hallucinatory episode.^147,150^ The basal forebrain is believed to be a key subcortical regulator of the DMN by governing the transition between internally and externally directed brain states.^199,200^ PD-associated VH may therefore relate to pathological alterations in cholinergic neurotransmission, which allows the DMN to exert excessive influence on perceptual processes. Consequently, a dysregulated DMN might prompt the release of self-generated or stored-memory content as a compensatory response in situations where external visual input is impoverished.

Attention

The allocation of voluntary attention has been frequently associated with increased cholinergic activity. Cortical acetylcholine promotes high frequency gamma oscillations within the PFC, which is critical for attentional and higher order cognitive processes and is believed to be dependent on α4β2 nAChRs.^201^ Indeed, PD patients with VH show a decrease in frontal gamma activity.^138^ This may be related to decreased nAChR density in the key attentional network regions, including the dorsolateral PFC and superior parietal lobe, presumably a consequence of degeneration of cholinergic neurons in the NBM.^81,189^

Attentional dysfunction in PD patients may contribute to VH through compromised precision control, whereby sensory evidence is not efficiently selected and enhanced. In this instance, impaired cholinergic signalling would result in a reduced level of precision assigned to parts of the visual processing streams, partially decoupling visual perception from external sensory input, which cannot then be recovered by attentional control.^173^ Cholinergic projections originating from both the PPN and NBM to the mediodorsal nucleus of the thalamus facilitate the reliability and efficacy of visual representation in cortical circuitry by maintaining active cortico-thalamo-cortical loops.^114,145,202^ In PD patients with VH, the mediodorsal thalamic nucleus is atrophic and shows substantial connectivity loss, which would potentially perturb signalling to the PFC.^53,100^ This may lead to impairments in attentional binding, allowing for inappropriate visual representations to be superimposed on otherwise veridical perceptual scenes.^145^

At the network-level, a hypocholinergic state resulting from degeneration of the ascending cholinergic systems may result in a failure to engage the DAN, allowing for internally generated information from unconstrained DMN activity to enter conscious awareness.^64,98^ The inability to effectively quiesce the DMN has been associated with nicotinic cholinergic signalling. This is evidenced in studies that found nAChR antagonists attenuate task-induced deactivation of DMN in healthy participants.^203^ Conversely, nAChR agonists cause reduced DMN activity with concomitant increases in activity in task-positive network hubs, as well as the insula and thalamus.^204^ The anterior insula exhibits reduced cholinergic activity and grey matter atrophy in PD patients with VH, which could restrict its ability to switch between attentional networks.^88,122^

Thalamic dysfunction may be linked with the network imbalance seen in PD patients with VH in addition to disrupting sensory gating of external and internal visual information. Reduced thalamic input from the cholinergic neurons in the PPN would potentially lead to hyperpolarization of the TRN, shifting thalamocortical neurons from tonic mode to burst-firing mode and promoting theta rhythms.^45,205^ The resulting thalamocortical dysrhythmia is believed to produce a state where the DMN is disinhibited, which in the early stages of disease may be corrected by re-engaging the attentional control networks.^150^ However, degeneration of the cholinergic projections from the NBM to the frontal and parietal cortex evolving with disease progression may reduce the ability to recruit these control networks.^64,98^ Impairments in attentional flexibility coupled with an overly influential DMN and compromised bottom-up perceptual processes may therefore underlie VH in PD.

Although we have evaluated the cholinergic contributions to each cognitive process in turn, it is clear they are inextricably linked, influencing perception dynamically and synergistically. These complex interactions are likely mediated by disruptions in distinct yet coordinated neuromodulatory systems, as well as alterations in thalamocortical communication, which ultimately leads to dysfunctional information flow. Unravelling the interdependence of these factors is therefore integral for the development of a comprehensive neurobiological framework that explains the mechanisms that drive the onset and maintenance of VH.

Implications for treatments targeting the cholinergic system

Given the established degeneration of cholinergic structures in PD, and its mechanistic role in VH discussed above, treatment focused on restoring cholinergic transmission would be expected to give benefit to neuropsychiatric symptoms. We discuss the relevance of the aforementioned mechanisms to the application of current and novel emerging treatments for VH targeting the cholinergic system.

Cholinesterase inhibitors

Several clinical trials have demonstrated the efficacy of ChEIs for addressing VH associated with PD and is first line treatment for VH in DLB.^13,206-210^ A recent meta-analysis revealed that PD-associated hallucinations and delusions were significantly attenuated by ChEIs, albeit with small effect sizes.^13^ A possible explanation for the modest effect sizes is that individual responses to ChEIs may depend on the baseline state of the system prior to treatment, potentially leading to an underestimation of their therapeutic benefit for the treatment of psychotic symptoms in PD.^86,211,212^ For example, there is evidence that mAChR integrity is associated with an increased response to the ChEI, donepezil, in both PDD^86^ and DLB.^211^ Furthermore, baseline NBM functional connectivity in MCI strongly predicts cognitive outcomes six months after treatment with ChEIs.^212^ Therefore, moving towards biomarker-driven therapeutic strategies might allow for the identification of individuals who are more likely to respond to indirect ‘cholinomimetic’ treatment.

Considering the relevance of cholinergic system degeneration in pathophysiological mechanisms that underscore VH in PD, investigating whether ChEIs can partially reverse changes in functional connectivity in treatment responders is of great interest. Although studies in PD are scarce, evidence from patients with AD has shown that treatment with AChE can increase within-network resting state functional connectivity in both the DMN^213^ and the DAN.^214^ It remains unclear whether increased coherence within these attentional networks might be related to symptomatic improvements following ChEI treatment in PD patients with VH or if systemic changes in network dynamics are possible. Future research investigating whole-brain functional network changes following ChEI treatment in PD patients will likely provide a better understanding of the underlying neurophysiological mechanisms responsible for remediating VH.

While the degeneration of cholinergic nuclei may be associated with a general dysregulation of cholinergic tone, neuroimaging studies in PD patients with VH suggest that cholinergic loss of function also results from the degeneration of the axons connecting these nuclei with cortical cholinergic receptors.^64,98^ This loss of synapses may provide an additional explanation regarding the variable outcomes of ChEIs in treating both hallucinations and cognitive impairment in PD. Therefore, in addition to inhibiting the degradation of acetylcholine which relies on the presence of intact cholinergic neurons, effective pharmacological intervention may also require direct stimulation of muscarinic and nicotinic receptors.

Receptor-specific agents

In contrast to the non-selective, tonic stimulation of cholinergic receptors associated with ChEIs, subtype-specific agents offer a targeted approach to improve cholinergic function in patients with PD. The development of highly specific cholinergic receptor agonists has faced challenges due to the conservation of the orthosteric site for acetylcholine binding.^215^ To overcome this hurdle, positive allosteric modulators (PAMs) have been a significant focus of drug discovery efforts as they exploit evolutionarily less conserved allosteric binding sites and therefore provide greater structural diversity for selective targeting.^216,217^ PAMs are distinct from direct agonists as they amplify the activation initiated by endogenous acetylcholine, preserving the native spatiotemporal patterns of cholinergic receptor activation.

Neuronal mAChRs represent viable targets for treating VH in PD given their relatively preserved densities compared with nAChRs, potentially related to their predominant postsynaptic location. Notably, the potent mAChR agonist xanomeline has been shown to reduce hallucinations and psychotic symptoms in AD^218^ and schizophrenia.^219^ Despite encouraging preliminary results, the further clinical development of xanomeline was hampered by adverse side effects, possibly due to the activation of peripheral receptors.^220^ Muscarinic receptor-knockout animal models revealed that the antipsychotic-like effects of xanomeline were primarily mediated by M1 and M4 mAChRs.^221^ Consequently, pharmacological investigations have focused on PAMs for these mAChR subtypes for the treatment of neuropsychiatric symptoms. Preclinical studies have highlighted the putative benefits of M1 and M4 selective PAMs.^222-225^ For instance, M1 PAMs, have been shown to restore attentional performance in rats with cholinergic lesions^222^ and improve memory deficits in murine models of PDD.^223^ Additionally, there is evidence that M4 PAMs exhibit antipsychotic properties, improve disruptions in sleep/wake architecture,^224^ and reduce levodopa-induced dyskinesias.^225^ Nicotinic receptors have received less attention given their potential for off-target effects. However, there are some PAMs for low-affinity α7 nAChRs, such as LL-00066471, that have shown potential to improve recognition memory and cognitive deficits in preclinical models of dementia.^226^

One concern regarding the efficacy of PAMs is that therapeutic effects would be dependent on the presence of sufficient levels of endogenous cholinergic activity. This would hinder their use as a monotherapy in the face of severe cholinergic deficits. There is some preliminary evidence that the action of PAMs can be further potentiated when combined with ChEIs,^227^ however this has yet to be explored in pathological models of LB disorders.

Nucleus basalis of Meynert deep brain stimulation

Deep brain stimulation (DBS) serves as a targeted method to selectively modulate neuronal activity within specific subcortical brain regions and is thought to induce network effects beyond the focal target.^228^ DBS delivered to either the subthalamic nuclei (STN) or internal globus pallidus (GPi) has reached standard-of-care status for its use as an advanced therapy to address motor symptoms in PD. Recently, attention has shifted towards novel neuromodulation targets, including the NBM, to tackle the cholinergic deficits associated with PDD and DLB.^229,230^

While the clinical outcomes of NBM-DBS have been variable and inconclusive, it appears to be relatively safe and well-tolerated in patients with Lewy body dementia, without substantial deterioration or serious adverse events.^229,230^ In a small randomized clinical trial involving six patients with PDD, bilateral NBM stimulation did not yield significant changes in cognitive impairment but led to an improvement in hallucinations and quality of life.^229^ This was particularly evident in patients experiencing daily complex VH despite prior treatment with ChEIs. Two of these patients reported a complete cessation of VH, suggesting that NBM stimulation might ameliorate treatment-refractory neuropsychiatric symptoms.

Currently, there is no strong consensus regarding the mechanisms that underscore the therapeutic outcomes of DBS. A general hypothesis is that electrical stimulation disrupts abnormal information flow through the stimulated nucleus to its synaptic targets.^231^ Unlike conventional DBS targets which aim to modulate activity within the cortico-basal ganglia-thalamic loop, stimulation of the NBM involves electrode placement within a degenerating nucleus with the goal to augment cholinergic activity. However, the paroxysmal nature of VH suggests that their presence may not necessarily be associated with a persistent hypocholinergic state, as might be the case for cognitive impairment, but rather dysregulated cholinergic neurotransmission leading to a transient disturbance of information processing. In this way, the signal for efficacy of NBM-DBS for the treatment of VH may be associated with the reinstatement of more physiological patterns of activity in the NBM.

Preclinical studies indicate that low-frequency stimulation of the NBM can promote the release of acetylcholine and enhance vasodilative responses, potentially improving neocortical functioning.^232-234^ Typically, clinical trials have employed continuous stimulation at 20 Hz, which is believed to be analogous to the natural firing rate of cholinergic neurons in the NBM.^232,235^ However, two recent studies in non-human primates revealed that intermittent stimulation increased the functional efficacy of the NBM-DBS, facilitating improvements in working memory and sustained attention.^236,237^ Considering the potential diverse impacts of different stimulation frequencies on local and global network dynamics, future research could explore alternative stimulation parameters targeting VH as the primary end point.^238^

Conclusion

Convergent evidence now suggests that VH in PD are related to dysfunction within a distributed neural system, whereby one or more components become either structurally or functionally pathological. This review has illuminated the dominant role of central cholinergic degeneration in contributing to these alterations and how it relates to the disrupted cognitive architecture that precipitates the hallucinatory perceptual experience. Beyond the cholinergic system, several other neurotransmitters have been implicated in the pathogenesis of VH. Exploring the interactions between these neuromodulatory systems holds promise for developing a comprehensive neurobiological framework. As cholinergic degeneration likely occurs before the onset of hallucinations, it will be interesting to investigate whether cholinomimetic pharmacotherapies for at risk patients may mitigate the progression to psychosis and further cognitive decline.^23^ Clinical research that leverages novel personalized modelling approaches may therefore assist in the development of treatment strategies tailored to individual patient needs.