Deciphering Freezing of Gait: What Neuropathology Reveals About an Episodic Phenomenon

ABSTRACT

Freezing of gait (FoG) is an episodic, debilitating motor phenomenon defined as paroxysmal episodes wherein there is an inability to step effectively, despite attempting to do so. Although FoG is most commonly associated with Parkinson's disease, it manifests across a spectrum of progressive neurodegenerative and potentially transient non‐neurodegenerative conditions that affect overlapping locomotor circuits. The episodic nature of FoG implies transient disruptions of a distributed locomotor network rather than fixed structural lesions, rendering its neuropathological basis particularly challenging to define. This review synthesizes current knowledge of the structural, neurochemical and proteinopathic substrates that precondition and precipitate FoG. The principal disease categories implicated in FoG are reviewed, including Lewy body diseases, progressive supranuclear palsy, multiple system atrophy, corticobasal degeneration and argyrophilic grain disease, as well as non‐neurodegenerative conditions such as cerebrovascular disease and metabolic encephalopathy. Mixed and overlapping pathologies are common. Across these disorders, different disease stages and phases, and their frequent combinations, interact with selective vulnerability of distinct neuronal and glial populations, giving rise to a complex pathophysiological landscape. It is proposed that this convergence results in heightened network susceptibility, whereby cumulative damage to cortical, subcortical and brainstem locomotor nodes progressively lowers the threshold for episodic circuit failure. Understanding these multilevel neuropathological determinants is essential for developing targeted therapeutic strategies aimed at stabilizing the locomotor network and reducing FoG burden.

Freezing of gait (FoG) occurs across Parkinson's disease, multiple neurodegenerative conditions and non‐neurodegenerative disorders. This review synthesizes the structural, neurochemical and proteinopathic substrates underlying FoG, showing how cumulative damage to distributed locomotor circuits—compounded by overlapping pathologies—progressively lowers the threshold for episodic circuit failure. Defining these neuropathological determinants is essential for developing targeted therapies to reduce FoG burden.

Abbreviations

4R

4‐repeat tau isoform

ABC

amyloid/Braak/CERad neuropathologic scoring system (NIA‐AA)

AD

Alzheimer's disease

ADNC

Alzheimer's disease neuropathological change

AGD

argyrophilic grain disease

APOE ε4

apolipoprotein E epsilon 4 allele

ARTAG

ageing‐related tau astrogliopathy

AT8

phospho‐tau antibody clone AT8

Aβ

amyloid‐beta

CA1

Cornu Ammonis 1 (hippocampal subfield)

CAA

cerebral amyloid angiopathy

CBD

corticobasal degeneration

CBS

corticobasal syndrome

CERAD

Consortium to Establish a Registry for Alzheimer's Disease

ChAT

choline acetyltransferase

CLN6

neuronal ceroid lipofuscinosis type 6

CPGs

central pattern generators

CSF

cerebrospinal fluid

CSVD

cerebral small vessel disease

FoG/FOG

freezing of gait

FTLD

frontotemporal lobar degeneration

GBA

glucocerebrosidase gene

GP

globus pallidus

ICFOG

International Consortium for Freezing of Gait

LATE

limbic‐predominant age‐related TDP‐43 encephalopathy

LATE‐NC

limbic‐predominant age‐related TDP‐43 encephalopathy neuropathological change

LBD

Lewy Body disease

LC

locus coeruleus

LDT

laterodorsal tegmental nucleus

MLR

midbrain locomotor region

MRI

magnetic resonance imaging

MSA

multiple system atrophy

MSA‐C

multiple system atrophy–cerebellar variant

MSA‐P

multiple system atrophy–Parkinsonian variant

NBM

nucleus basalis of Meynert

NCS

nucleus centralis superior

NFT

neurofibrillary tangle

NIA‐AA

National Institute on Aging–Alzheimer's Association

NPH

idiopathic normal pressure hydrocephalus

NROP

nucleus raphe obscurus‐pallidus

PAG

periqueductal grey

PAGF

pure akinesia with gait freezing

PART

primary age‐related tauopathy

PD

Parkinson's disease

PMRF

pontomedullary reticular formation

PNLA

pallido‐nigro‐luysian atrophy

PPFG

primary progressive freezing gait

PPN

pedunculopontine nucleus

PSP

progressive supranuclear palsy

PSP‐CBS

PSP–corticobasal syndrome variant

PSP‐F

PSP–frontal variant

PSP‐P

PSP–Parkinsonism phenotype

PSP‐PGF

PSP with progressive gait freezing

PSP‐SL

PSP–speech/language variant

TDP‐43

TAR DNA‐binding protein 43

TpOH

tryptophan hydroxylase

1. Introduction

Freezing of gait (FoG) is one of the most disabling and enigmatic manifestations of a neurological phenomenon. The 2025 ICFOG definition describes FOG as “paroxysmal episodes wherein there is an inability to step effectively, despite attempting to do so” (Gilat et al. 2026). While its prevalence in advanced Parkinson's disease (PD) has long been recognized, FoG is increasingly recognized in other conditions, arising whenever specific nodes of the locomotor network are sufficiently compromised. Unlike fixed motor deficits, FoG fluctuates in relation to environmental triggers, attentional demands, emotional stress and pharmacological state (Nutt et al. 2011). This implies that FoG does not reflect irreversible structural destruction alone but rather represents a dynamic collapse of network synchrony in a brain rendered susceptible by underlying structural damage. The concept of ‘locomotor network vulnerability’ provides the conceptual framework for understanding why FoG emerges episodically rather than persistently (Weiss et al. 2020).

From a neuropathological perspective, this framework presents a fundamental challenge. Neuropathology excels at identifying fixed structural lesions and protein aggregates, yet FoG, by definition, reflects transient dysfunction rather than permanent ablation. Establishing causal neuropathological correlates therefore requires distinguishing lesions that precondition the network, creating vulnerability, from those that are merely coincidental findings. This distinction is rarely straightforward, and the field has not yet resolved whether specific regional pathologies are causally linked to FoG, or whether the network as a whole must reach a threshold of cumulative damage before episodic freezing can occur. Indeed, the heterogeneous risk profile of FoG underscores the multidimensional nature of pathophysiology and indicates that no single neurotransmitter system or anatomical structure accounts for its emergence.

This review examines the neuropathological basis of FoG from two complementary angles: first, by surveying the diseases—both neurodegenerative and non‐neurodegenerative—that damage brain regions relevant to FoG; second, by exploring how the combination of diseases leads to the degeneration of the locomotor network and its multiple neurotransmitter systems creates the conditions for episodic gait failure. The review concludes by integrating these structural, proteinopathic and neurochemical mechanisms into a unified model through the concepts of mixed pathology and network susceptibility.

2. Neuroanatomy and Neurotransmitter Systems Underlying FoG

Effective gait requires the dynamic integration of cortical executive control, subcortical motor programming and automatic brainstem and spinal rhythm generation (Nutt et al. 2011). The locomotor network responsible for this integration is anatomically distributed and functionally hierarchical. Cortical signals descend through corticostriatal and corticoreticular pathways to engage the basal ganglia loop, the midbrain locomotor region (MLR), comprising the pedunculopontine nucleus (PPN), cuneiform and subcuneiform nucleus, and the pontomedullary reticular formation (PMRF), which ultimately drives spinal central pattern generators (CPGs) responsible for rhythmic locomotor output (Nutt et al. 2011). Based on functional neuroimaging studies, walking engages a distributed network of cortical regions involved in motor planning, execution, sensory integration and visuospatial processing. In addition, cortical areas such as the supplementary motor area and lateral premotor cortex, and further brain regions including medial primary sensorimotor cortex, cingulate cortex, parahippocampal cortex, superior parietal lobe and occipital cortex are also considered (Nutt et al. 2011).

Neuroimaging studies reveal that FoG arises from distributed network dysfunction rather than a single lesion site (Fasano et al. 2015). Atrophy and hypometabolism affecting frontal regions impair motor planning and executive control, while parietal and posterior cortical regions show changes linked to deficits in visuospatial processing and navigation. Subcortically, the caudate nucleus, putamen and thalamus demonstrate denervation and volume loss, disrupting movement initiation and sensorimotor relay. At the brainstem level, atrophy and reduced connectivity compromise gait initiation and locomotor control, and cerebellar under‐activation further impairs rhythm and balance (Fasano et al. 2015). Indeed, based on lesion network mapping, lesions that cause FoG converge on a shared functional network connected to the cerebellar locomotor region (Fasano et al. 2017). Collectively, FoG reflects a failure of connectivity across cortical executive‐attention networks and subcortical locomotor circuits, with this widespread network disruption underlying its characteristic episodic presentation. Furthermore, comorbid depression and anxiety as risk factors in FoG (Weiss et al. 2020) underscore the importance of limbic system involvement, particularly the amygdala, in linking neuropathological changes to this episodic motor phenomenon. Finally, experimental models show that lesions of the ventral periaqueductal grey (PAG) attenuate freezing behaviour, whereas lesions of the dorsolateral PAG exacerbate freezing (De Oca et al. 1998).

Regarding neurotransmitters, the basal ganglia, receiving dopaminergic input from the substantia nigra, modulate cortical motor programs through direct and indirect pathway signalling. Dopaminergic depletion, the hallmark of PD, disrupts this modulation and is the principal target of current pharmacological therapy. However, the incomplete and often paradoxical responsiveness of FoG to levodopa indicates that the dopaminergic system alone cannot account for its pathophysiology (Nutt et al. 2011). Cholinergic neurons of the PPN, particularly the pars compacta, play a key role in gait initiation and postural control, while the pars dissipata also contains glutamatergic, GABAergic and monoaminergic neurons. Glutamate provides the main excitatory drive from the cortex and brainstem locomotor regions to spinal CPGs. In contrast, GABAergic output from the globus pallidus interna tonically inhibits the mesencephalic locomotor region, with disinhibition promoting locomotion. Noradrenaline and serotonin further modulate descending brainstem–spinal pathways, supporting gait rhythm, tone and adaptability. FoG thus reflects a systems‐level failure of this multi‐transmitter architecture rather than an isolated dopaminergic deficit.

Another neurological phenomenon is festination. Festination is distinct from FoG but shows partial overlap: Type I festination (often termed ‘festinating‐freezing’) is closely related to FoG and, by consensus, is considered a manifestation of FoG, whereas Type II festination represents a separate, balance‐related gait disturbance and should be classified as a festination‐related impairment rather than FoG (Gilat et al. 2026). Although neuropathological evidence on festination is limited, neuroimaging studies suggest that cognitive motor control, postural regulation, memory and emotional factors appear to be closely linked to festination in PD (Lee et al. 2025). The absence of a clear association with striatal dopaminergic activity points to the contribution of extranigral mechanisms beyond primary dopaminergic dysfunction (Lee et al. 2025).

3. Diseases Affecting Brain Regions Relevant to FoG

FoG has been reported in a range of conditions associated with both neurodegenerative and non‐neurodegenerative causes (Figure 1).

FIGURE 1.

Classification of diseases contributing to freezing of gait. Conditions are divided into two broad categories based on their underlying aetiology and potential for reversibility. Non‐neurodegenerative causes (green, left) represent potentially treatable and reversible contributors, detectable primarily through neuroimaging studies. Neurodegenerative causes (red, right) represent progressive and non‐reversible contributors, characterized primarily through neuropathology studies. A central mixed pathology category (blue ellipse) acknowledges that combinations of the above, with or without Alzheimer's disease neuropathological change (ADNC). The brain schematic illustrates relevant neural circuit involvement, with red connections indicating disrupted pathways associated with neurodegenerative conditions and green connections representing circuits implicated in non‐neurodegenerative aetiologies. Created in BioRender. Kovacs, G. (2026).

3.1. Non‐Neurodegenerative Causes

A clinically important category of FoG‐associated pathology comprises non‐neurodegenerative conditions that structurally or functionally compromise locomotor circuitry. These conditions are of particular relevance because they may produce FoG‐like disturbances in the absence of neurodegeneration and because some are at least partially reversible, providing natural experiments for understanding the anatomical requirements for FoG generation.

Cerebrovascular disease is among the documented non‐neurodegenerative causes of FoG. Ischemic strokes involving bilateral PPN (Kuo et al. 2008) can produce gait freezing likely by directly interrupting the MLR. Three patients with isolated dorsal midbrain infarcts presented with prominent gait hesitation and irregular, ataxic gait without parkinsonism or peripheral cerebellar dysfunction; all had associated oculomotor abnormalities, including third nerve palsies and/or vertical gaze deficits (Hathout and Bhidayasiri 2005). A haemorrhage at the pontomesencephalic junction involving the right pedunculopontine area rendered an 83‐year‐old woman unable to stand or initiate stepping movements (Masdeu et al. 1994). Lacunar infarcts in the basal ganglia, thalamus or periventricular white matter can similarly disrupt cortico‐subcortical locomotor loops without producing the classical parkinsonism of PD. Similarly, cortical venous thrombosis has been reported to lead to reversible parkinsonism, altered consciousness and progressive gait slowing (Jenkins et al. 2001). Right hemorrhagic stroke of the supplementary motor area was also associated with injury with gait apraxia including FoG (Yada and Kawasaki 2022). Furthermore, chronic subdural hematoma (Noda et al. 2017) represents additional vascular mechanisms capable of producing episodic or persistent gait dysfunction by compromising cortical motor and supplementary motor areas or their efferent pathways.

Idiopathic normal pressure hydrocephalus (NPH) constitutes another cause, in which gait disturbance arises from distortion of periventricular frontal pathways subserving locomotor control (Giladi et al. 1997). The reversibility of gait impairment with cerebrospinal fluid diversion in NPH underscores the functional nature of the underlying network disruption in these cases. Metabolic encephalopathies, including haemodialysis (Nakajima et al. 2013) and liver disease‐associated (Sousa et al. 2019) metabolic syndrome, represent a further category, which can produce transient gait abnormalities without fixed structural lesions. While the former was thought to be associated with vasogenic edema on bilateral basal ganglia associated with posterior reversible encephalopathy syndrome, in the latter, the role of ammonia was discussed. Additional non‐neurodegenerative causes documented in neuroimaging and clinical series include neoplastic lesions (metastatic disease and lymphoma), carbon monoxide toxicity (Lee et al. 2010) that preferably damages the globus pallidus, and post‐surgical states affecting basal ganglia or brainstem structures (Nutt et al. 2011).

Although the clinical course and symptom constellations differ in these conditions from FoG associated with neurodegenerative diseases, some similarity in gait disturbances supports the notion that these brain regions are critical to the pathogenesis of FoG. These non‐neurodegenerative aetiologies demonstrate that FoG can arise from focal or diffuse disruption of specific network nodes without underlying neurodegeneration or neuronal loss, directly validating the network vulnerability model. They also carry important diagnostic implications: FoG of vascular, metabolic or structural origin may respond to entirely different interventions than those effective in neurodegenerative conditions such as PD, making etiological distinction a clinically meaningful step in management.

3.2. Neurodegenerative Diseases: Fundamentals and Relevance for FoG

Neurodegenerative diseases are clinically and pathologically diverse conditions characterized by progressive neuronal dysfunction and loss, commonly beginning with synaptic disruption before overt cell death becomes apparent (Forrest and Kovacs 2025). The modern conceptual framework of ‘proteinopathy’ recognizes that each major neurodegenerative disease is defined by the characteristic accumulation of specific misfolded proteins in filamentous aggregates within neurons, astrocytes or oligodendroglia. These proteinopathic deposits exhibit distinct biochemical properties, structural strains, stereotypical anatomical staging patterns and associations with specific clinical phenotypes (Forrest and Kovacs 2025).

The pathogenesis of these disorders involves misfolding of specific proteins that can propagate through connected brain regions in staged patterns, acquire toxic properties, and form aggregates that sequester other proteins, impair proteostasis, disrupt RNA processing, associate with loss‐of‐function of proteins, and activate neuroinflammatory cascades (Forrest and Kovacs 2025). Classification of neurodegenerative diseases reflects their presumed aetiologies (idiopathic, genetic or acquired), their anatomical patterns of neuronal loss and protein deposition, and the biochemical and structural properties of their predominant misfolded proteins (Forrest and Kovacs 2025). Understanding these levels of classification is essential when interpreting how specific pathologies affect FoG‐relevant circuitry.

Neurodegenerative disorders associated with FoG are predominantly from the group of diseases classified as α‐synucleinopathies and tauopathies. Within the α‐synucleinopathies, PD and multiple system atrophy (MSA) are of primary relevance, whereas the tauopathies most commonly linked to FoG include progressive supranuclear palsy (PSP) and corticobasal degeneration (CBD). Among these conditions, PD and PSP are most frequently associated with FoG, reflecting their prominent involvement in basal ganglia–brainstem locomotor networks. FoG has been reported in less common neurodegenerative diseases, including the tauopathy argyrophilic grain disease (AGD) and certain lysosomal storage disorders, such as neuronal ceroid lipofuscinosis.

3.3. The Concept of Mixed Pathologies

Neurodegenerative diseases have traditionally been classified by their predominant misfolded protein; however, neuropathological and biomarker evidence increasingly challenges this single‐protein paradigm (Forrest and Kovacs 2023). Many individuals simultaneously harbour multiple misfolded proteins—including amyloid‐β, tau, α‐synuclein and TDP‐43—forming an individualized ‘protein code’ that shapes prognosis and therapeutic response in ways that single‐target models cannot capture (Kovacs et al. 2010). These co‐pathologies are not merely incidental: they interact, influencing each other's aggregation kinetics and sequestering soluble, physiologically active protein forms, thereby altering their availability and detectability in cerebrospinal fluid and plasma. Where deposits overlap anatomically, such interactions may modify the clinical phenotype beyond what any single proteinopathy would produce in isolation. Cerebrovascular disease and systemic disorders add a further layer of complexity when they co‐occur with proteinopathies (Forrest and Kovacs 2023). Capturing this biological reality requires integrating biomarkers of protein aggregation with markers of neuroinflammation, neuronal injury, genetic risk and vascular pathology—stratifying patients by their combined pathological burden rather than their dominant clinical diagnosis (Kovacs et al. 2010). This approach holds the potential to explain past trial failures, identify distinct progression trajectories and guide the development of personalized, combination‐aware therapies.

4. Major Neurodegenerative Diseases Affecting FoG‐Relevant Brain Regions

4.1. PD and Lewy Body Diseases (LBDs)

PD is the paradigmatic and most extensively studied neurodegenerative disease associated with FoG. It is characterized by the progressive aggregation of α‐synuclein into Lewy bodies and Lewy neurites, spreading in a stereotypical caudo‐rostral pattern (Braak et al. 2003). PD belongs to the broader group of LBD—which also includes PD dementia and dementia with Lewy bodies (Fu and Halliday 2025)—all of which share a common α‐synuclein filament fold structure, as revealed by cryogenic electron microscopy (Yang et al. 2022). Braak staging captures the sequential anatomical progression of this pathology: early stages are confined to the olfactory bulb, dorsal vagal nucleus and lower brainstem; intermediate stages involve the substantia nigra pars compacta, locus coeruleus (LC) and raphe nuclei; and later stages reach limbic and neocortical regions integral to gait control and executive function (Braak et al. 2003).

Recent studies challenge the notion of a uniform pattern of LBD pathology to include body‐first and brain‐first patterns and extend its cytopathological features beyond neurons (Andersen et al. 2025). Notably, many antibodies used in early neuropathological studies failed to detect α‐synuclein pathology in glial cells. Subsequent methodological advances have revealed prominent astroglial α‐synuclein pathology, particularly in the striatum from Braak stage 3 onward, as well as in the amygdala and cortical regions at later stages (Kovacs et al. 2014; Altay et al. 2022). In addition, bulky oligodendroglial inclusions are observed in Braak Stage 4 and advanced disease, most prominently within the pallido‐thalamic tract and also in the substantia nigra (Rohan et al. 2016). Together, these findings underscore a significant role for glial involvement in PD pathogenesis and highlight glial dysfunction as a potentially relevant contributor to the mechanisms underlying FoG. These aspects have recently been integrated into a staging system that accounts for protein pathology across different affected cell types (Otero‐Jimenez et al. 2025).

The clinical hallmarks of PD, bradykinesia, rigidity and tremor, principally reflect dopaminergic degeneration of the substantia nigra. FoG is more associated with the non‐tremor dominant forms of PD (Weiss et al. 2020). The tremor‐dominant form of PD differs, for example, from the akinetic rigid form most notably in the regional pattern of substantia nigra degeneration, the latter showing greater overall nigral cell loss concentrated ventrolaterally, while the former shows less total loss but more medial nigral and retrorubral A8 damage (Boonstra et al. 2022). At a broader level, akinetic rigid form carries heavier cortical Lewy body and amyloid burden with more diverse co‐pathologies (Boonstra et al. 2022). Molecularly, widespread neuronal mitochondrial complex I deficiency strongly predicts non‐tremor dominant phenotypes, suggesting the two subtypes reflect fundamentally different neurodegenerative processes rather than different disease stages (Flones et al. 2024).

FoG in PD correlates more strongly with extranigral pathology (Weiss et al. 2020). This multi‐system involvement explains the frequent incompleteness of levodopa responses in FoG and supports combination therapeutic approaches targeting non‐dopaminergic transmitter systems. Importantly, FoG appears later in the disease course in PD, associated with higher Hoehn and Yahr stage and longer duration of levodopa treatment (Giladi et al. 1992). Risk factors for the future development of FoG in PD span motor, non‐motor, imaging and genetic domains. Non‐modifiable factors include older age at disease onset, longer disease duration, lower dopamine transporter uptake in the caudate and putamen, presence of the APOE ε4 allele, and reduced CSF Aβ42 levels (Herman et al. 2023). Among modifiable factors, greater motor severity—particularly the postural instability and gait difficulty subtype—alongside depression, anxiety, cognitive impairment and higher levodopa equivalent daily doses are the strongest predictors (Herman et al. 2023). Non‐motor features, including autonomic dysfunction, sleep disturbance and hyposmia, further contribute to risk, while neuroimaging evidence of white matter hyperintensities increases FoG risk more than threefold (Herman et al. 2023). Collectively, this multidimensional risk profile reinforces the view that FoG emerges from the convergence of multiple pathological processes rather than any single causal factor. It must be noted that LBD can be rarely the morphological substrate of pure akinesia with gait freezing (PAGF) (Williams et al. 2007), which is mostly associated with PSP pathology. Interestingly, the case reported in that study (Williams et al. 2007) had also occasional neuronal and glial tau inclusions in the pontine base, STN, striatum, thalamus and substantia nigra, which would be currently suggestive of a concomitant early‐stage PSP pathology (Kovacs et al. 2020).

4.2. MSA

MSA is an α‐synucleinopathy that is pathologically distinguished from LBDs by the predominant accumulation of α‐synuclein in oligodendroglial cytoplasmic inclusions—termed Papp‐Lantos bodies (Gilman et al. 2008). However, neuronal cytoplasmic and nuclear inclusions constitute an important feature of the disease (Wiseman et al. 2025). MSA presents in two principal clinical subtypes: MSA‐P (parkinsonian variant), in which striatonigral degeneration dominates, and MSA‐C (cerebellar variant), in which olivopontocerebellar atrophy is the principal substrate. A lesser‐known variant, atypical MSA (also termed frontotemporal lobar degeneration‐synuclein), presents initially with frontotemporal dementia‐like symptoms before evolving into classical MSA clinical features (Aoki et al. 2015). Neuropathologically, it is distinguished by prominent neuronal α‐synuclein pathology in the limbic system superimposed on the classical oligodendroglial inclusions of subcortical regions (Rohan et al. 2015). Notably, while the α‐synuclein filament fold in typical MSA differs from that seen in LBDs (Schweighauser et al. 2020), atypical MSA harbours a hybrid Lewy–MSA fold specifically within the limbic system, suggesting that variability of cellular protein pathologies, filament folds, can be associated with distinct clinical phenotypes (Enomoto et al. 2025).

In MSA‐P, degeneration of putamen, caudate, globus pallidus and substantia nigra disrupts the corticostriatal and nigrostriatal circuits central to locomotor programming. α‐Synuclein pathology is thought to begin simultaneously across the striatum, lentiform nucleus, substantia nigra, brainstem white matter tracts, cerebellar subcortical white matter, and motor, midfrontal and sensory cortices (Phase 1), subsequently spreading to the spinal cord and thalamus (Phase 2), then to the hippocampus and amygdala (Phase 3), and finally reaching the visual cortex (Phase 4) (Brettschneider et al. 2018).

In MSA‐C, cerebellar and brainstem involvement—including pontine nuclei, inferior olivary complex and cerebellar Purkinje cells—disrupts the cerebellar contribution to gait. A‐synuclein pathology is thought to progress through four sequential phases: beginning in the cerebellum and cerebellar‐brainstem connections (Phase 1), spreading to pyramidal and extrapyramidal white matter (Phase 2), then involving neocortical and basal ganglia grey matter (Phase 3) and finally reaching the amygdala and hippocampus (Phase 4) (Brettschneider et al. 2017).

FoG is considered an additional feature in MSA (Poewe et al. 2022). In a clinicopathological study of 15 patients with pathologically confirmed MSA, the median age at onset was 56 years and median survival was about 6.5 years. FoG was present in 40% of patients at the first clinical visit and increased to 54% by the final visit, indicating progression with disease duration (Muller et al. 2002). Restricted neuropathological examination did not identify a substrate for FoG. In other pathology‐based studies, FoG was not reported as a leading symptom (Wenning et al. 1997).

4.3. PSP

PSP is a main form of tauopathies characterized by the accumulation of misfolded tau composed of 4‐repeat (4R) isoforms of tau. The Movement Disorders Society criteria (Hoglinger et al. 2017) distinguish eight clinical phenotypes, Richardson syndrome being the most frequent. Further subcortical phenotypes such as PSP‐parkinsonism (PSP‐P) and PSP with progressive gait freezing (PSP‐PGF) follow a slower course with longer survival, whereas cortical phenotypes—including PSP‐frontal (PSP‐F), PSP‐speech/language variant (PSP‐SL), and PSP‐corticobasal syndrome (PSP‐CBS)—present with more rapid decline.

Misfolded tau accumulates in characteristic morphological forms—subcortical neurofibrillary tangles (NFTs), tufted astrocytes and oligodendroglial coiled bodies—following a hierarchical staging pattern that preferentially and early affects subcortical and brainstem nuclei integral to locomotion (Kovacs et al. 2020). Early‐stage PSP pathology is prominent in the globus pallidus, subthalamic nucleus and striatum, with subsequent extension to frontal cortex and dentate nucleus of the cerebellum, and eventual involvement of occipital regions in the most advanced cases (Kovacs et al. 2020). Studies have shown intra‐disease variability in PSP, supported by recent findings of differences in high–molecular weight tau assemblies and seeding activity across cases (Martinez‐Valbuena et al. 2025). PSP clinical variants further demonstrate heterogeneity in cortical vulnerability and in the cellular composition of tau pathology, involving neurons, astrocytes or oligodendrocytes (Kovacs et al. 2020).

Neuronal tau pathology in the midbrain, including the substantia nigra, superior colliculus and PAG, and in the subthalamic nucleus is particularly relevant to FoG in PSP. These structures constitute critical nodes of the locomotor network: the subthalamic nucleus modulates basal ganglia output via the indirect pathway, while midbrain structures govern postural reflexes and vertical gaze control. PSP also demonstrates a profound multi‐transmitter profile likely contributing to the early onset, severity and levodopa‐unresponsiveness of FoG and postural instability in PSP.

While FoG is frequently present as a symptom in PSP, the concept of primary progressive freezing gait (PPFG), introduced in 1993 (Achiron et al. 1993), along with the term gait ignition failure (Atchison et al. 1993), and reported in earlier studies as isolated or pure akinesia (summarized by Compta et al. 2007), requires special attention. Finally, the 2017 Movement Disorders Society criteria recognized progressive gait freezing (PSP‐PGF) as a distinct clinical phenotype (Hoglinger et al. 2017). Based on the criteria, probable PSP‐PGF requires progressive gait freezing within 3 years plus either vertical supranuclear gaze palsy or slow vertical saccades, while possible PSP‐PGF requires progressive gait freezing within 3 years alone, without the ocular motor features. While there are a few neuropathological studies on PSP‐PGF, systematic mapping of different cytopathologies as compared with all other clinical phenotypes are lacking. One study describing six cases of PAGF reported a milder degree of PSP pathology and noted several atypical pathological features (Williams et al. 2007). In particular, the distribution of pathology differed from that seen in classic PSP, with relative sparing of the pontine base. These aligns with the unpublished experience of the author of the present article, while ADNC as co‐pathology might be present in some cases. Apart from PAGF, an earlier study on a cohort of 24 patients with PSP, FoG was present in 25% at initial evaluation and increased to 53% by the final visit, becoming more common with disease progression (Muller et al. 2002).

Similar patterns of restricted pathology have been reported in cases with pallido–nigro‐luysian atrophy (PNLA); however, not all such cases present with primary FoG, and many exhibit other clinical features of PSP (Ahmed et al. 2008). More recently, a small cohort of patients with PSP with a markedly protracted disease course has been described, characterized by long intervals between major disease milestones, although FoG was not the dominant clinical feature (Couto et al. 2022). Neuropathologically, these cases showed only very mild alterations along the pallido–nigro‐luysian axis, with mild to moderate neuronal loss, suggesting a relatively benign molecular behaviour of misfolded tau. These cases differ from previously reported PSP phenotypes with long disease duration (Lukic et al. 2022), where cytopathological differences have been observed, but not in the restricted pattern described in protracted‐course PSP (Couto et al. 2022). Together, these observations suggest that early tau molecular behaviour and selective vulnerability of specific microcircuits may strongly influence both clinical presentation and disease progression. Introduction of a 4R tau seeding amplification assay from cervical skin will facilitate in vivo diagnosis of PSP (Martinez‐Valbuena et al. 2024), which, together with biomarker possibilities to exclude synuclein (Martinez‐Valbuena et al. 2026) and ADNC type co‐pathologies, will facilitate better understanding of disease progression and pathogenesis in clinical cohorts.

4.4. CBD

CBD is also a major 4R tauopathy, distinguished neuropathologically from PSP by its characteristic astrocytic plaques, tau‐positive threads and coiled bodies within white matter, alongside cortical neuronal loss with ballooned neurons and tau inclusions in the frontoparietal cortex. CBD is therefore a neuropathological diagnosis. CBS refers to a characteristic clinical presentation that may be associated with CBD pathology (Saranza et al. 2019). The clinical phenotype is determined by the regional distribution of neuronal loss and tau lesions. Consequently, CBD pathology can give rise to a broad spectrum of clinical manifestations, and conversely, CBS may result from diverse underlying neuropathologies (Saranza et al. 2019). CBD shares with PSP an early predilection for basal ganglia and brainstem nuclei, disrupting motor circuits, but its greater cortical involvement—particularly of the motor and premotor cortex, supplementary motor area and parietal regions—contributes to the additional clinical features of apraxia, alien limb phenomenon and cortical sensory loss (Saranza et al. 2019).

Studies of incidental CBD cases suggest that cortical astrocytic tau pathology precedes neuronal tau accumulation, representing an early pathogenic event (Ling et al. 2016). This is supported by symptomatic CBD cases in which astrocytic tau deposition occurs in regions lacking neuronal tau pathology (Kovacs et al. 2018). CBD shows a staged progression of astrocytic tau pathology, spreading from frontal–parietal to temporal–occipital cortex, then to subcortical regions and brainstem, accompanied by intracellular maturation from fine granular deposits in distal astrocytic processes to more advanced tau accumulations (Kovacs et al. 2017). A staging system that integrates neuronal and glial tau pathology and clinical phenotypic variability has not yet been developed. The neuropathological tau load in CBD correlates with disease severity and progression. Emerging evidence points to tau strain heterogeneity (Sanders et al. 2014) as a potential determinant of regional and cytopathological variability, exemplified by the predominance of neuronal tau pathology in the malignant form of CBD (Ling et al. 2020). While gait disturbance and postural instability are recognized features of CBD, the specific neuropathological determinants of FoG in this condition remain less thoroughly characterized than in PSP or PD, partly because CBD is rarer and its clinical and pathological heterogeneity is considerable. In a cohort of 13 patients with CBD, FoG was uncommon, affecting 8% at first presentation and 25% by the final visit, remaining less frequent than in other atypical parkinsonian disorders. This low prevalence may partly reflect the diagnostic difficulty in distinguishing FoG from apraxia, and no consistent pathological correlate was identified (Muller et al. 2002).

4.5. Differential Neurotransmitter Degeneration in Parkinsonian Disorders: Insights From Postmortem Studies Relevant to FoG

While developments in neuroimaging clearly advanced the spectrum of neurotransmitter alterations (Kanel et al. 2022), it is worth returning to archival literature on postmortem studies that collectively paint a picture of overlapping yet disease‐specific neurodegeneration across the neurotransmitter systems in parkinsonian disorders.

Regarding the serotonergic raphe system, a clear dissociation between oral and caudal raphe nuclei was demonstrated (Kovacs et al. 2003): tryptophan hydroxylase‐positive (TpOH+) neurons were significantly reduced in the nucleus centralis superior (NCS) in PD, pointing to oral raphe vulnerability, while the nucleus raphe obscurus‐pallidus (NROP) showed a paradoxical increase in TpOH+ neuron proportion in PSP and MSA, suggesting either compensatory upregulation or preferential loss of non‐serotonergic neurons in the caudal raphe (Kovacs et al. 2003). Preservation of the oral raphe was confirmed by others focusing on the dorsal raphe nucleus (Eser et al. 2018). The increased proportion of TpOH+ neurons in the NROP in PSP and MSA most likely reflects preferential loss of non‐serotonergic raphe neurons rather than a true increase in serotonergic cells, shifting the balance toward serotonergic dominance in this nucleus (Kovacs et al. 2003). Since the NROP projects descending pathways preferentially to axial motoneurons and participates in locomotor rhythm production and supraspinal reflex modulation, this imbalance might eventually contribute to the axial rigidity and postural instability that define both conditions. The disproportionate serotonergic drive onto axial motor pools, normally counterbalanced by non‐serotonergic raphe neurons, may produce dysregulated tonic motor facilitation, potentially underpinning freezing episodes and gait dysfunction. Notably, PSP and MSA share a common pattern: both exhibits disproportionately heavier protein pathology in non‐raphe brainstem regions, independent of whether the aggregating protein is tau or α‐synuclein (Kovacs et al. 2003). It remains unclear, however, whether protein pathology selectively involves the non‐serotonergic cells in the NROP. Similarly to this investigation, MSA stood out strikingly in another study, which found sparing of rostral raphe serotonergic neurons (dorsal and median raphe) despite severe cholinergic depletion (Benarroch et al. 2002), contrasting with PD, where rostral raphe neuron loss has been documented.

The cholinergic PPN emerges as the most consistently vulnerable structure across diseases, but with differing severity. It was demonstrated that PPN neurons were reduced to approximately 43% of control values in PD and to 21% in PSP (Hirsch et al. 1987), establishing PSP as having the more severe and consistent cholinergic brainstem deficit. These findings were corroborated in another study reporting over 50% neuronal loss in PD and approximately 60% in PSP, with NFTs affecting 40%–64% of remaining PPN neurons in PSP versus bodies in 6%–39% in PD (Jellinger 1988). Another study indicated glutamatergic neuronal loss also in the PPN both in PSP and CBD with higher proportion of neurons with tau pathology (Eser et al. 2018). Notably, ad showed comparatively mild PPN loss (~34%), with tangles in fewer cells, suggesting relative cholinergic brainstem sparing. In MSA severe PPN depletion (to ~25% of controls) alongside equally severe loss in the laterodorsal tegmental nucleus (LDT), a pattern broadly similar to PD but previously undocumented in MSA was reported (Benarroch et al. 2002). CBD, examined by showed a distinct profile (Kasashima and Oda 2003): the LDT was largely preserved, while only mild PPN neuronal reduction was found, in sharp contrast to PSP, which exhibited severe LDT cholinergic loss (~41% reduction in ChAT‐positive neurons) alongside high NFT burden—making LDT vulnerability a potential neuropathological discriminator between PSP and CBD. Across the basal forebrain, the nucleus basalis of Meynert (NBM) was more severely affected in a single case studied CBD than in PSP (Kasashima and Oda 2003). Additional striatal cholinergic deficits were also highlighted (Warren et al. 2005). Importantly, despite all this cholinergic depletion, consistent failure of cholinesterase inhibitors in clinical trials for this disease emphasize further aspects to be explored (Warren et al. 2005).

The LC, the brain's principal source of noradrenaline, is differentially affected across FoG‐relevant neurodegenerative syndromes. In PD, the LC undergoes noradrenergic neuron loss and Lewy body accumulation (Braak et al. 2003), peaking further in PD with dementia (Fructuoso et al. 2025). In MSA, LC degeneration is also severe and is accompanied by glial cytoplasmic inclusions (Benarroch et al. 2002; Campese et al. 2021); uniquely in MSA, noradrenergic degeneration extends beyond the LC to include near‐complete loss (~90%) of the pontine A5 noradrenergic group, which projects to sympathetic preganglionic neurons (Benarroch et al. 2002). In PSP, LC neurodegeneration is driven by tau accumulation, with an average 49% reduction of pigmented noradrenergic neurons relative to controls and 44% of surviving neurons bearing tau‐positive inclusions; importantly, the degree of LC cell loss correlates with clinical severity of PSP even after adjusting for disease duration, implicating noradrenergic deficits in the heterogeneous clinical expression of the syndrome (Kaalund et al. 2020). CBD shares with PSP a pattern of prominent LC tau burden and neurodegeneration (Ohm et al. 2020). These findings contrast with another study that reported increased tau pathology but relatively preserved neuronal numbers in PSP and CBD compared with ad (Eser et al. 2018).

In summary, PSP is characterized by severe, widespread cholinergic neuronal loss spanning the PPN (including glutamatergic cells as well), LDT, NBM, striatum, and brainstem eye movement nuclei, with relative serotonergic predominance over non‐serotonergic neurons (including GABAergic, glutamatergic, or neuropeptidergic populations) in the caudal raphe projecting toward the spinal cord; PD shows moderate and variable PPN/LDT loss, oral raphe serotonergic depletion; MSA uniquely combines severe mesopontine cholinergic depletion with striking serotonergic raphe preservation; and CBD is distinguished from PSP by greater forebrain (NBM) cholinergic loss with relative LDT sparing, though the limited case numbers in CBD studies temper firm conclusions. Importantly, the LC is a convergent site of pathology across these disorders, yet the underlying protein, the severity and spatial extent of noradrenergic loss, and the clinical correlates differ.

A major limitation of postmortem neurotransmitter studies is the lack of clinically homogeneous phenotypes across cohorts, the small number of cases examined in some studies, the use of heterogeneous methodologies and the failure to assess all key nuclei within the same studies, which hampers interpretation of neurotransmitter alterations within individual cases. Innovative approaches to investigate the whole extent of these nuclei as shown for the LC are needed (Gilvesy et al. 2022).

4.6. Regional and Cellular Distribution of Protein Pathology

Systematic mapping of the stage‐dependent regional vulnerability and cell‐type‐specific burden of protein pathology across PD, MSA‐P, MSA‐C, PSP and CBD highlights fundamental differences in both the sequential staging and cellular tropism of pathology across parkinsonian syndromes. PD follows a caudal‐to‐rostral neuronal progression, MSA is distinguished by early and severe oligodendrocytic involvement across multiple regions, PSP by early and widespread neuronal and astrocytic burden with prominent subcortical predilection, and CBD by a uniquely pan‐cellular, high‐burden pattern that does not conform to a linear staging model—reflecting the biological distinctiveness underlying their overlapping clinical presentations.

Regarding FoG‐relevant brain regions a complex pattern emerges when comparing these disorders (Figure 2). Across neurodegenerative diseases, brain regions are affected at different times and involve different cell types. In PD, the frontal cortex is involved relatively late and mainly shows neuronal pathology, whereas in MSA, especially in MSA‐P, this region is affected earlier in contrast to MSA‐C. PSP engages the frontal cortex at an intermediate stage, with a strong predominance of astrocytic pathology alongside neuronal and oligodendrocytic involvement, a pattern similar to CBD, where all three cell types are heavily affected although no specific staging is defined. The striatum is an early and severely affected region in MSA‐P due to marked oligodendrocytic pathology, while in PD it is involved in the fourth out of six stages and affects both neurons and astrocytes. PSP affects the striatum early, again with prominent astrocytic pathology, whereas CBD shows consistently high pathology across neurons, astrocytes and oligodendrocytes. The globus pallidus is particularly vulnerable in PSP and CBD, showing severe neuronal pathology and mostly in PSP, oligodendroglial involvement; in contrast, PD affects this region late and moderately, and MSA‐P affects it very early but largely through oligodendrocytes. The subthalamic nucleus is a key distinguishing site for PSP, where it is affected very early and severely in neurons, while it is spared in MSA and less affected in PD. In the midbrain tegmentum, PD shows variable neuronal involvement reflecting disease heterogeneity, whereas MSA‐P and PSP show earlier involvement, but PSP shows clearly more prominent mixed neuronal‐glial pathology compared with MSA‐P where oligodendrocytes can be affected; in MSA‐C, this region is involved in a later phase. In particular, the PPN and the PAG are predominantly affected in PSP and PD; early in PSP, and in PD from the transition between Braak Stages 3 and 4 (Braak et al. 2003). The ponto‐medullary reticular formation is the earliest affected region in PD, with severe neuronal pathology, while MSA‐P, PSP and CBD show intermediate involvement. The cerebellum is especially vulnerable in MSA, particularly MSA‐C, with early oligodendrocytic pathology, whereas PD shows only mild and late involvement. Involvement of the dentate nucleus (neurons) and cerebellar white matter (oligodendrocytes) in PSP is accumulating from intermediate disease stages to prominent levels. Finally, the spinal cord is affected at early‐intermediate stages across disorders, with neuronal pathology predominating in PD, mixed glial involvement in MSA and PSP, and high pathology in CBD.

FIGURE 2.

Regional distribution and cellular pathology of α‐synuclein and tau inclusions across parkinsonian disorders. Schematic summary comparing the burden of pathological protein aggregates across eight brain regions thought to be relevant for freezing of gait in five neurodegenerative conditions: Parkinson's disease (PD), multiple system atrophy with predominant parkinsonism (MSA‐P), multiple system atrophy with predominant cerebellar ataxia (MSA‐C), progressive supranuclear palsy (PSP), and corticobasal degeneration (CBD). Brain regions examined include the frontal cortex (Fr Cx), striatum (Str), globus pallidus (GP), subthalamic nucleus (STN), cerebellum (CbII), midbrain (MB), pons, and spinal cord (SpC), each illustrated by a corresponding neuroanatomical diagram with the region of interest highlighted in dark blue. For each region and disease, pathological inclusion burden is shown separately for three cell types: neurons, astroglia, and oligodendroglia. Semiquantitative scores indicate absent (−), minimal (−/+), mild (+), moderate (++), or severe (+++). While heterogeneity between patients is acknowledged, these scores reflect the general trend of cytopathological involvement in these disorders. Background shading indicates the disease stage or phase at which pathology is typically observed: red = early stage; orange = intermediate stage; yellow = late stage; grey = data not available. While disease stages and phases vary (e.g., PD and PSP, 6 stages; MSA‐P and MSA‐C, 4 phases), these were compressed into three groups for comparability. Column header shading denotes the predominant protein involved: teal/cyan = α‐synuclein pathology (PD, MSA‐P, MSA‐C); light blue = tau pathology (PSP, CBD). Created in BioRender. Kovacs, G. (2026).

4.7. Less Common Neurodegenerative Diseases Associated With FoG

In genetic PD, GBA is the strongest and best‐evidenced risk gene for FOG. In a longitudinal study, GBA carriers developed FOG significantly earlier than idiopathic PD (Yang et al. 2023). Two mechanisms appear operative: an Aβ‐dependent pathway, in which reduced CSF Aβ42 in GBA carriers predicts FOG and mediates gait impairment via cognitive decline, and an Aβ‐independent pathway likely driven by α‐synuclein accumulation (Yang et al. 2023). By contrast, parkin‐associated PD shows FOG as a late complication, emerging after decades of disease (Doherty et al. 2013; Barbosa et al. 2025). Clinicopathological data indicate restricted ventral nigral degeneration with relative sparing of cognitive and brainstem nuclei, implying that symptomatology in parkin is driven mainly by ventral nigral degeneration rather than by α‐synuclein spread or amyloid pathology (Doherty et al. 2013; Barbosa et al. 2025).

AGD is a 4R‐tauopathy characterized by early atrophy of the ambient gyrus and abundant argyrophilic, AT8‐positive tau grains predominantly affecting medial temporal lobe structures (Tolnay and Clavaguera 2004) and a distinct filament fold (Shi et al. 2021). Core neuropathological features include tau‐positive grains, pretangles, oligodendroglial coiled bodies and granular/fuzzy astrocytes—most prominently in the amygdala—with relatively few senile plaques and low Braak NFT stages (I–III) (Tolnay and Clavaguera 2004). AGD pathology shows a stereotyped, stage‐dependent progression from the ambient gyrus and CA1 to limbic, paralimbic and eventually neocortical regions (Saito et al. 2004), and is commonly observed as a comorbid pathology in the elderly, where it may lower the threshold for dementia.

AGD usually presents with amnestic syndrome and neuropsychiatric symptoms, but a recent case report described an 80‐year‐old man in whom AGD extension into the midbrain and substantia nigra produced a clinical picture indistinguishable from PSP, including gait freezing, postural instability and bradykinesia over an 11‐year course (Sengoku et al. 2025). Autopsy with Western blotting and immunoelectron microscopy confirmed AGD‐specific tau pathology distinct from PSP. Authors concluded that clinicians should consider AGD in elderly patients with parkinsonism and cognitive decline, particularly when features such as early memory disturbance, slow progression, right‐dominant limbic atrophy on imaging and normal CSF tau levels are present.

Neuronal ceroid lipofuscinosis type 6 (CLN6) is a rare autosomal recessive lysosomal storage disorder causing progressive neurodegeneration through intraneuronal ceroid lipopigment accumulation, typically presenting between 18 months and 8 years with seizures, vision loss, psychomotor regression, myoclonus and ataxia, leading to rapid decline and early death. A recent study documented an 8‐year‐old boy who developed FoG at age 4 as an isolated early symptom, preceding all other motor signs and occurring without classic parkinsonian features such as bradykinesia, tremor or rigidity (Damasio et al. 2014). MRI findings of white matter changes and atrophy in the frontal lobe, internal capsules and pons support the authors' hypothesis that storage material disruption of frontal‐pontine gait circuits underlies the FoG (Damasio et al. 2014).

4.8. Co‐Pathologies Affecting Brain Regions Relevant for FoG

Regarding mixed pathologies relevant to FoG, several conditions warrant consideration (Figure 3), including Alzheimer's disease neuropathologic change (adNC), limbic‐predominant age‐related TDP‐43 encephalopathy neuropathologic change (LATE‐NC), AGD, ageing‐related tau astrogliopathy (ARTAG), primary‐age‐related tauopathy (PART), cerebrovascular disease and metabolic encephalopathy. LBD should also be acknowledged as a common co‐pathology; regarding FoG, it is important to note that LBD has been reported in up to 20% of cases with PSP and CBD, and in approximately 5% of cases with MSA.

FIGURE 3.

Overview of FoG‐relevant neurodegenerative conditions, co‐pathologies, and affected brain regions. The figure illustrates the relationships between primary neurodegenerative conditions associated with Freezing of Gait (FoG), their frequent co‐pathologies, and the brain regions predominantly affected. Left panel: FoG‐relevant primary conditions include Lewy Body Disease (LBD), Multiple System Atrophy (MSA), Progressive Supranuclear Palsy (PSP), and Corticobasal Degeneration (CBD), each characterized by distinct misfolded protein aggregates (colour‐coded by protein type). Blue arrows indicate the primary condition–FoG association; red arrows indicate co‐pathology contributions converging on a shared pathological hub. Middle panel: FoG‐relevant co‐pathologies are grouped by their principal misfolded protein: Alzheimer's Disease Neuropathological Change (ADNC; Aβ and Tau), Primary Age‐Related Tauopathy (PART; Tau), Argyrophilic Grain Disease (AGD; Tau), Age‐Related Tau Astrogliopathy (ARTAG; Tau), Limbic‐predominant Age‐related TDP‐43 Encephalopathy Neuropathological Change (LATE‐NC; TDP‐43), Lewy Body Disease (LBD; α‐Synuclein), and Cerebral Small Vessel Disease (CSVD). Right panel: Co‐pathologies differentially impact cortical (yellow), limbic (purple), and subcortical‐brainstem (grey) brain regions, as depicted on the medial brain schematic, reflecting the topographic distribution of pathology relevant to FoG pathophysiology. Colour legend—Misfolded proteins: Pink = Aβ and Tau; Orange‐red = Tau; Cyan = TDP‐43; Green = α‐Synuclein. Brain regions: Yellow = Cortical; Purple = Limbic; Grey = Subcortical‐Brainstem (Bst). Abbreviations: FoG, Freezing of Gait; LBD, Lewy Body Disease; MSA, Multiple System Atrophy; PSP, Progressive Supranuclear Palsy; CBD, Corticobasal Degeneration; ADNC, Alzheimer's Disease Neuropathological Change; PART, Primary Age‐Related Tauopathy; AGD, Argyrophilic Grain Disease; ARTAG, Age‐Related Tau Astrogliopathy; LATE‐NC, Limbic‐predominant Age‐related TDP‐43 Encephalopathy Neuropathological Change; CSVD, Cerebral Small Vessel Disease; Aβ, Amyloid‐beta; TDP‐43, TAR DNA‐binding protein 43; Bst, Brainstem.

adNC is defined by extracellular accumulation of amyloid‐β (Aβ) fibrils and intraneuronal deposition of abnormally phosphorylated tau as NFTs (Forrest and Kovacs 2025). Both pathologies show stereotyped spatial progression. Aβ deposition advances from neocortical regions (phase 1) to the hippocampus and entorhinal cortex (2), basal ganglia and diencephalon (3), brainstem (4) and cerebellum (5). Cerebral amyloid angiopathy (CAA) frequently accompanies parenchymal Aβ deposits and is classified as capillary (type 1) or non‐capillary (type 2), each with progressive stages. Traditional CERad criteria focus on argyrophilic neuritic plaques, representing only a subset of Aβ morphologies. Importantly, isolated Aβ deposition without tau pathology is insufficient for an ad diagnosis, although it is now viewed as preclinical ad when detected by biomarkers or post‐mortem in cognitively unimpaired individuals. Tau pathology follows the Braak sequence, beginning in the transentorhinal cortex (stage I), entorhinal cortex (II), inferior and middle temporal gyri (III–IV) and later neocortical regions, including occipital cortex (V–VI). Evidence of early tau involvement in the brainstem has led to proposed pre‐cortical staging, supporting the concept that AD pathology begins decades before symptoms. Based on NFT distribution, clinicopathological subtypes (e.g., typical, limbic‐predominant, hippocampal‐sparing) have been described. Neuropathological diagnosis currently follows NIA‐AA guidelines using the ABC system, integrating Aβ phase (A), tau Braak stage (B) and neuritic plaque density (C) to classify AD neuropathologic change as none, low, intermediate or high (Montine et al. 2012).

Recent clinicopathological evidence shows that ADNC significantly worsens FoG in PD (Parmera et al. 2025). A higher global AD pathology burden—especially amyloid‐β deposition—is associated with earlier onset and greater severity of FoG, whereas tau pathology shows no independent effect (Parmera et al. 2025). The findings suggest a synergistic interaction between amyloid‐β and α‐synuclein, possibly through cholinergic pathway degeneration. Further frequent forms of co‐pathologies as listed below have not yet been evaluated systematically to have an effect on FoG.

PART is a common age‐associated tauopathy observed in individuals over 60 years of age (Crary et al. 2014). It is defined by NFT pathology limited to Braak stages ≤IV, with prominent hippocampal involvement, few or no amyloid‐β plaques, and therefore does not meet criteria for adNC. It occurs in neurologically normal individuals as well as in those with cognitive impairment and encompasses NFT‐predominant dementia or tangle‐only associated cognitive impairment.

LATE‐NC is a recently defined, common TDP‐43 proteinopathy strongly associated with advanced age (typically > 80 years) and cognitive impairment, occurring without a clinical motor neuron disease or frontotemporal dementia phenotype (Nelson et al. 2019). It is characterized by TDP‐43–immunoreactive neuronal cytoplasmic inclusions and/or dystrophic neurites, predominantly affecting the medial temporal lobe. Although limbic TDP‐43 pathology has long been recognized in aging and various neurodegenerative disorders, LATE‐NC is now considered distinguishable from FTLD‐TDP and may occur in isolation, though it frequently coexists with Alzheimer's disease (Nelson et al. 2019). In the neuropathologic practice, a three‐tiered staging is used, emphasizing early involvement of the amygdala, followed by the hippocampus and frontal lobe (Nelson et al. 2023). Hippocampal sclerosis commonly accompanies LATE‐NC and is reported in up to 20% of individuals over 85 years. It is defined by disproportionate neuronal loss and gliosis in the hippocampal CA1 region and subiculum and is strongly associated with TDP‐43 deposition.

As discussed above, AGD is a limbic tauopathy that often presents as co‐pathology. AGD stages II or III were present in 18.5% of LBD cases, but no differences were found in the frequency of AGD among the LBD subgroups (Colom‐Cadena et al. 2017), it can be present in approximately 60% of PSP and practically all CBD cases.

ARTAG is a tau astrocytic pathology commonly observed in individuals over 60 years of age. Consensus criteria define five ARTAG types—grey matter, white matter, perivascular, subpial and subependymal—characterized by thorn‐shaped and/or granular fuzzy astrocytes (Kovacs et al. 2016). ARTAG can occur in isolation but more often coexists with neurodegenerative diseases. Predilection sites include the medial temporal lobe (especially the amygdala) and basal forebrain, with disease‐specific distribution patterns (Kovacs et al. 2017). Different ARTAG types associate with distinct clinicopathological features and aging measures, though their clinical significance remains incompletely understood (Kovacs 2020). Importantly, ARTAG is thought to develop in response to barrier dysfunction and to reflect the cumulative accumulation of eventually transient pathological events (Kovacs 2020).

Among non‐neurodegenerative co‐pathologies, small vessel disease is common and typically manifests as fibrohyalinosis of small vessels in the basal ganglia and cerebral lobar white matter. Even in the absence of a definitive infarct, enlargement of perivascular spaces can increase reactive astrogliosis in the white matter and basal ganglia. In addition, fibrotic changes in small vessels may cause chronic cerebral hypoperfusion, affecting neural circuits relevant for gait. These effects can be transiently exacerbated by fluctuations in blood pressure or metabolic disturbances, such as changes in blood glucose levels.

In addition to vascular disease, metabolic encephalopathy—even at low or subclinical levels related to liver or kidney disease—is common and potentially reversible. These conditions arise from direct neurotoxic effects (e.g., ammonia, urea), impaired cerebral energy metabolism, systemic inflammation and disrupted excitatory–inhibitory neurotransmission and are associated with the development of so‐called metabolic astroglia (Alzheimer type II astrocytes), underscoring the central role of astrocytes in pathogenesis (Gelpi et al. 2020).

Regarding the latter two as co‐pathologies, an interesting observation from a longitudinal aging study described individuals initially classified as having clinically possible AD who later showed cognitive improvement, giving rise to the impression of reversible cognitive decline (Kovacs et al. 2013). In these cases, vascular lesions or metabolic encephalopathy were proposed as contributing factors, occurring in the context of a low‐grade neurodegenerative disorder. This observation may be relevant for understanding how metabolic states or transient alterations in cerebral perfusion can precipitate or exacerbate neurological symptoms.

5. Toward an Integrated Neuropathological Framework to Understand FoG

FoG likely arises from impaired gait pattern generation, disrupted central drive, reduced movement automaticity, abnormal posture‐gait coupling, perceptual dysfunction and deficits in frontal executive control (Nutt et al. 2011). Although the anatomical structures involved have been generally defined, the mechanisms underlying transient neuronal dysfunction at the cellular level remain poorly understood. Understanding neurodegeneration requires moving beyond isolated protein pathology to a systems‐level framework that accounts for disease stage, co‐pathologies (vascular, neuroinflammatory, metabolic) and proteinopathy subtypes (Figure 4). This reframes diagnosis from a categorical label into a dynamic, individualized pathological map that better reflects heterogeneity, informs prognosis and enables targeted intervention. FoG is best understood not as the product of a single lesion but as an emergent failure of a locomotor network operating near its functional threshold—a ‘network vulnerability’ model (Weiss et al. 2020) in which cumulative structural and neurochemical compromise erodes normal compensatory capacity. As this reserve is depleted, physiological fluctuations in attention, emotion, neurotransmitter tone or environmental complexity can precipitate episodic gait collapse (Nutt et al. 2011; Weiss et al. 2020). This implies that observed neuropathological abnormalities may represent a heterogeneous mix of direct causes, permissive preconditions, incidental comorbidities or secondary consequences—a complexity that must inform how imaging and post‐mortem findings are interpreted. Indeed, the broad, heterogeneous clinical risk profile associated with FoG in PD (Herman et al. 2023) supports the notion here that low‐level co‐pathologies precondition individuals to states that aggravate or precipitate freezing episodes.

FIGURE 4.

Integrated neuropathology‐based conceptual framework for the pathophysiology of Freezing of Gait (FoG). The model in the upper panel illustrates how cumulative neurodegeneration and co‐pathologies lower the functional threshold for FoG episodes. Risk factors contribute to progressive neurodegeneration across motor, perceptual, cognitive, and emotional brain networks (represented by increasingly affected brain sections, highlighted in yellow to red). Combinations of neurodegenerative co‐pathologies with limbic predominance, acting as preconditioning factors, compound network susceptibility. Transient triggers (e.g., environmental or situational stressors) act upon this preconditioned state to precipitate an acute FoG episode when network function falls below the functional threshold (indicated by the red dashed line). Non‐neurodegenerative co‐pathologies (hepatic, renal, and vascular, represented by organ icons) further contribute to overall network vulnerability. Astroglia, which normally serve as a neuron support hub, undergo dysfunction under these conditions (shown by the transition from a healthy to a dysfunctional astrocyte morphology), amplifying network susceptibility. The integrated concept in the lower panel summarizes that a FoG episode results from the convergence of four key elements: underlying neurodegeneration, co‐pathologies (both neurodegenerative and non‐neurodegenerative), astrocytic dysfunction, and transient triggers. Together, these factors combine to precipitate a clinical FoG episode, characterized by a sudden, transient inability to initiate or continue gait. Created in BioRender. Kovacs, G. (2026).

Although there are relatively few observations on how co‐pathologies influence the development of FoG, several considerations can be drawn from known anatomical–pathological patterns. First, some co‐pathologies involve brain regions critical for gait control and may therefore predispose the system to transient failures that result in FoG. Importantly, ADNC involves brainstem nuclei with neurotransmitter relevance to FoG (Eser et al. 2018; Gilvesy et al. 2022). Second, the consistent involvement of the limbic system—particularly the frequent and severe involvement of the amygdala—suggests an indirect effect on gait. This aligns with clinical observations that emotional factors, including anxiety and depression, can contribute to the development or exacerbation of FoG. Indeed, the limbic system, particularly the amygdala, appears to play a critical role in FoG by driving heightened striato‐limbic connectivity that overwhelms top–down attentional control (Gilat et al. 2018). This dysfunctional fronto‐striato‐limbic interaction, likely rooted in anxiety and fear of falling, may precipitate gait freezing when the additional cognitive demands of walking further challenge an already overburdened system (Gilat et al. 2018). Beyond this, the amygdala is implicated in vestibular processing relevant to FoG (Bohnen et al. 2022), reflecting the extensive integration of vestibular and limbic networks that enable vestibular stimulation to modulate emotional behaviour via higher central and autonomic nervous system pathways (Rajagopalan et al. 2017). For the sake of completeness, it should be noted that the PAG, which is considered as an extralimbic anxiety (‘fear conditioning’) area (De Oca et al. 1998), is affected early in PSP and in intermediate stages of PD. Furthermore, beyond the amygdala, the co‐pathological involvement of the hippocampus is particularly significant given its central role in multisensory processing (Raccah et al. 2026) and spatial navigation (Strange et al. 2014), both of which are critical to the development of FoG (Fearon et al. 2015; How et al. 2022). Third, as observed in the primary disorders associated with FoG (i.e., PD, PSP, CBD), astroglial responses play a central role in certain co‐pathologies such as cerebral response to systemic metabolic conditions. Given their functions in buffering neurotoxicity and regulating ion transport, astrocytes are particularly relevant when considering mechanisms of transient neural dysfunction.

6. Conclusion: Contribution of Neuropathology

In conclusion, neuropathological investigation advances understanding of FoG by identifying lesions in gait‐control regions, uncovering low‐burden combined pathologies concentrated in the limbic system, alongside primary neurodegenerative disease that compound neuronal dysfunction in gait‐relevant brain regions, which collectively predispose to freezing, and by implicating astroglia as an underrecognized contributor to episodic neurological dysfunction. Postmortem studies have been particularly instrumental in delineating further the neurotransmitter spectrum of nuclei implicated in FoG. Nevertheless, research on the neuropathological basis of FoG remains limited, largely due to its intermittent nature and the lack of well‐characterized prospective cases (Weiss et al. 2020). The inherently cross‐sectional nature of postmortem investigation further prevents dynamic tracking of how neurochemical and structural changes evolve in relation to FoG onset and progression. Progress will therefore require multimodal, in vivo approaches interrogating distributed neural networks across neurotransmitter systems (Weiss et al. 2020) into a more comprehensive mechanistic framework. Moving forward, neuropathological inquiry must also expand beyond protein aggregation characterization to encompass astroglial reactions bridging the gap between static lesion confirmation and a dynamic understanding of FoG. Finally, rigorous studies examining the contribution of co‐pathologies lay the groundwork for multimodal biomarker approaches in clinical practice (Martinez‐Valbuena et al. 2026), with the aim of improving prediction of FoG across disease populations. Together with further research into shifts in neurotransmitter balance, these efforts may ultimately enable more effective therapeutic strategies.

Author Contributions

Gabor G. Kovacs: conceptualization, writing – original draft, writing – review and editing, visualization.

Funding

This work was supported by the Rossy Family Foundation.

Conflicts of Interest

GGK reports personal fees from Ferrer, Parexel, Treventis, and Mitsubishi‐Tanabe; funding from Rossy Family Foundation, from Edmond Safra Foundation, grants from Krembil Foundation, MSA Coalition, MJ Fox Foundation, Parkinson Canada, NIH, Canada Foundation for Innovation, and Ontario Research Fund outside the submitted work; in addition, GGK has a shared patent for 5G4 Synuclein antibody and a pending patent for Diagnostic assays for movement disorders (18/537,455), and received royalties from Wiley, Cambridge, Taylor and Francis, and Elsevier publishers. None of these had influence on the preparation of this manuscript.

Kovacs, G. G. 2026. “Deciphering Freezing of Gait: What Neuropathology Reveals About an Episodic Phenomenon.” European Journal of Neuroscience 63, no. 12: e70587. 10.1111/ejn.70587.

Data Availability Statement

No data were generated for the research described in this review article.