Parkinson's disease (PD), one of the most common neurodegenerative diseases with a rising global prevalence (Wolff et al., 2023), is caused by the widespread deposition of misfolded α-synuclein (αSyn), a neuronal protein forming intraneuronal Lewy bodies (LBs) and Lewy neurites, the morphological hallmarks of PD and related synucleinopathies. These changes cause progressive degeneration not only of the dopaminergic striatonigral system, responsible for the core motor symptoms, but of many other neuronal systems and organs. The resulting biochemical deficits are responsible for the heterogeneous spectrum of motor and non-motor symptoms of PD that contribute to the overall disease burden of this multisystem/multiorgan disorder (Dickson et al., 2009; Jellinger, 2012, 2022; Poewe et al., 2017). Its complex pathogenic pathways including oxidative stress, abnormal proteasome management, mitochondrial, synapse and lysosomal dysfunction, neuroinflammation, and other molecular processes, related to spreading of abnormal proteins in cortical and subcortical regions with subsequent dysregulation of multiple neuromodulator systems, are well-documented (Zaman et al., 2021; Koros et al., 2022; Calabresi et al., 2023; Hussein et al., 2023; Ye et al., 2023).

Cognitive impairment (CI) has been increasingly recognized as an integral and most disabling non-motor symptom of PD that shows a broad spectrum from subjective cognitive decline (SCD) or subjective cognitive complaints (SCCs) and mild cognitive impairment (MCI) to full-blown Parkinson's disease dementia (PDD). They can be present decades before manifest motor symptoms, i.e., in the pre-clinical phase of PD (Jessen et al., 2014; Weil et al., 2018; Flores-Torres et al., 2023). SCD in PD is a self-perceived decline in cognitive abilities with normal age-, sex-, and education-adjusted performance on standardized cognitive tests indicating adequate cognitive function (Oedekoven et al., 2022), although there is currently no consensus of clinical criteria for SCD in PD (Oedekoven et al., 2022; Huang J. et al., 2023). Since it is already associated with metabolic changes in multiple cortical regions indicating early aberrant pathological changes (Ophey et al., 2022), SCD is considered an increasing risk of developing dementia (Pike et al., 2022). SCCs refer to self-perceived cognitive decline in cognitive normal individuals and are considered a preclinical sign of subsequent CI. SCCs as measured by the Cognitive Complaints Interview (CCI) are strongly correlated with objective cognitive performance; the CCI score increases with age and is inversely correlated with cognitive performance (Hong et al., 2018). Since there are no definite guidelines for PD-SCD (and -SCCs), researchers have classified it differently, and some defined it as subjective CI without objective cognitive decline but poor performance in action naming (Galtier et al., 2023) and an indicator of subsequent CI (Purri et al., 2020; Hong and Lee, 2023), although they are very common in PD and increase with duration of PD (Hong et al., 2018). Inclusion of SCCs in PD-MCI criteria may strengthen the ability to detect persons with risk factor for future cognitive decline in early PD (Jones et al., 2023), although the prevalence and risk factors for SCCs have been found to be distinct in PD with and without MCI, suggesting that SCCs in early PD with different cognitive status may have different pathogenicity (Pan et al., 2021).

The concept of MCI, described in Prichard (1837), encompasses an intermediate stage between normal cognitive ability and dementia, in which SCD and objective CI are present, while activities of daily life are preserved (Petersen et al., 1999; Albert et al., 2011). MCI, representing early clinical features of cognitive disorders, such as Alzheimer's disease (AD) or other dementia disorders (Morris et al., 2001; Winblad et al., 2004; Gauthier et al., 2006; Jicha et al., 2010), describes a gradual decline of cognitive abilities affecting single or multiple cognitive domains on complex tasks (Petersen et al., 2009; Litvan et al., 2012; Pedersen et al., 2017). MCI affecting newly diagnosed untreated (de novo) and juvenile PD patients may be associated with subtle changes of cognitive function that are not recognized by patients, family members or even clinicians (Bougea et al., 2019). These changes may occur in pre-symptomatic stages of PD (Fengler et al., 2017), preceding the onset of severe CI/dementia by up to 20 years, as well as at the time of diagnosis or during the whole disease process. PD-MCI has an increased risk for conversion to PDD (Janvin et al., 2006; Williams-Gray et al., 2007; Litvan et al., 2012; Nicoletti et al., 2019; Giil and Aarsland, 2020; Wallace et al., 2022). The heterogeneity of PD-MCI may reflect the underlying pathogenic mechanisms as well as the impact of co-morbidities, such as AD-related changes, that are incompletely understood. The present article is intended to review the currently available clinical, epidemiological, neuroimaging data and biomarkers concerning PD-MCI, based on literature research in PubMed and Google Scholar until January 2024 and personal experience in a large number of autopsy-proven PD cases with and without cognitive dysfunctions.

PD-MCI has a high variability in severity, rate of progression and type of affected cognitive domains (Aarsland et al., 2021; Koros et al., 2022; Carceles-Cordon et al., 2023). Furthermore, there is a heterogeneity of cognitive symptoms in various PD subtypes, patients with non-tremor dominant motor symptoms and with depression developing more often and more severe MCI (Tremblay et al., 2013). Although it can manifest as deficits in nearly all cognitive domains, executive dysfunction is most pronounced (Zgaljardic et al., 2003; Kudlicka et al., 2011) that may manifest earlier and more frequent than previously appreciated (Tröster, 2008). The Movement Disorder Society Task Force (MDS-TF) has delineated diagnostic criteria for PD-MCI (Litvan et al., 2012; Goldman et al., 2015). They are as follows: (1) a diagnosis of PD, (2) a gradual decline of cognitive ability in the level I category based on a brief cognitive assessment or on level II category based on a more comprehensive neuropsychological assessment including at least two tests for each of five cognitive domains (attention, executive function/EF/, language memory, visuospatial function, and memory), and (3) complete functional independence. Impairment in neuropsychological tests is defined per one to two standard deviations below the appropriate norms, a significant decrease in the serial cognitive tests, or a significant decrease from the estimated premorbid levels. PD-MCI patients according to the level II criteria show impairment in at least two tests within one single cognitive domain or different domains. Exclusion criteria are definite PDD, other explanations for Cl and PD-associated co-morbidities. Other definitions for PD-MCI are impairment of at least one cognitive domain, defined as 1.5 times worse performance on z-scores, compared with normal controls (Caviness et al., 2007; Aarsland et al., 2010), and deficits in at least four cognitive domains (executive, attention, visuospatial abilities, and memory), being severe enough to interfere with routine functions in everyday live (Svenningsson et al., 2012; Kiesmann et al., 2013). For other criteria see Saredakis et al. (2019). The most frequent phenotypes of MCI in prodromal PD are deficiencies in executive, visuospatial and linguistic tasks and multi-domain amnestic phenotypes (Ciafone et al., 2020), without essential affection of memory (Speelberg et al., 2022). PD-MCI subjects were divided into four subgroups: non-amnestic multiple-domain type (PD-naMCI-MuD), non-amnestic single-domain type (PD-naMCI-SiD), amnestic single-domain type (PD-aMCI-SiD), and amnestic multiple-domain type (PD-aMCI-MuD). Differences between PD-MCI and PD-NC with respect to age, age at onset, years of education, and motor symptom severity were significant. In contrast to the population as a whole, where aMCI is the most common subtype, in PD naMCI is dominant, with prominent executive and attention dysfunctions (Yarnall et al., 2013), whereas in another group PD-aMCI was most common (Vasconcellos et al., 2019). In a large US sample, the most common subtype was PD-naMCI-SiD (47.7%), followed by aMCI-MuD (24.2%), aMCI-SiD (18.8%), and naMCI-MuD (9.5%; Goldman et al., 2012a). In others, the single-domain type was the largest PD-MCI subgroup (52.83%), and within single domain impairment, non-amnestic is more common than the amnestic one (Litvan et al., 2011). Another study included 39.4% naMCI-SD, 30.5% aMCI-MuD, 23.4% naMCI-MuD, and 6.7% aMCI-SiD subtypes, EFs being most frequently impaired (Kalbe et al., 2016). A recent meta-analysis confirmed a higher frequency of multiple domain MCI subtype (Baiano et al., 2020). The most common PD-aMCI subtype was associated with a lower quality of life compared with the non-amnestic group (Vasconcellos et al., 2019). Memory and EF impairment were most frequent (22.64 and 20.75%, respectively). The overall cognitive function in aMCI-MuD was significantly worse compared with the SiD type (Nie et al., 2019). These lesions were more pronounced in early stages of PD, executive dysfunction being most pronounced. Working memory appeared to decline slightly faster in early PD compared to healthy controls (HCs), while other domains remained similar without considerable changes (Turner et al., 2023). PD-MCI patients had greater subjective cognitive and communicative problems than PD-NC ones (Jaramillo-Jimenez et al., 2023). Basic PD-MCI in addition to diffuse impairment of EF, involved visuospatial ability and construction, language, working memory, slowed processing speed and verbal fluency, as well as impaired attention, daytime sleepiness (Ciafone et al., 2020). Clear deficits in EF and attention, observed 6 years before diagnosis, were predictive of poorer outcomes at follow-up (Williams-Gray et al., 2007). A diagnosis of prodromal PD should be considered in individuals with MCI who present with prominent executive, attentional and visuospatial deficits, worse verbal fluency and less memory impairment (Yoon et al., 2014; Brønnick et al., 2016). A meta-analysis of five longitudinal studies of PD-MCI reported a predominant involvement of visuospatial skills and visual memory as well as longitudinal executive dysfunction (Wallace et al., 2022). Heterogeneous cognitive subgroups warrant further investigation to advance effective evaluation (Yang et al., 2023).

PD-MCI patients show slower non-decision times than controls, which is negatively correlated with visuospatial performance, but they have no impairment of information processing ability. Pausing before word production is associated with PD-MCI, and these pauses are correlated with Montreal Cognitive Assessment (MoCA) score but not with motor severity (Andrade et al., 2023). Moreover, PD-MCI is associated with gait dysfunction (Morris et al., 2017), characterized by changes in mean velocity, step and circle length and other features (Amboni et al., 2022; Russo et al., 2023). Other changes in individuals with PD and MCI are sleep problems and reduced proprioceptive and olfactory sensitivities, which are related to disease severity (Li et al., 2023).

MCI is common in PD but heterogeneous, with cognitive subtypes, but the underlying pathobiological mechanisms have not been fully understood. A combined analysis of voxel based morphometry (VBM) evaluating the regional changes combined with functional connectivity was used to explore the relevant morphological and functional changes and their association with CI in PD, but the findings are not always concordant (Hou and Shang, 2022).

Whole-brain analyses revealed increased global brain atrophy (Mak et al., 2017) or mild diffuse brain atrophy (Martin et al., 2009; Pereira et al., 2014). For major neuroimaging findings in PD-MCI see Supplementary Table 1.

PD patients with normal cognition at baseline (PD-NC) who develop MCI during follow-up are classified as “PD-MCI converters,” while those PD-MCI patients who became asymptomatic during the follow-up are referred to as “non-converters.” PD-MCI converters have a reduced GM volume of temporal regions and the amygdala-hippocampus network at baseline, associated with early right temporal atrophy, progressive frontal lobe atrophy and WM volume reduction (Wen et al., 2015; Zhou et al., 2020; Chen et al., 2021). Others described GM thinning in anterior cingulate, temporal, parietal and occipital cortices, progressive atrophy of the caudate nucleus, thalamus, nucleus acccumbens and CA2/3 hippocampal subregion (Foo et al., 2017; Filippi et al., 2020; Gorges et al., 2020).

VBM analysis of PD-MCI converters showed lower GM density in the left prefrontal areas, left insular cortex, and bilateral caudate nucleus, and smaller substantia innominata volume compared with non-converters (Lee et al., 2014). Subcortical shape analysis revealed smaller volumes in the bilateral thalamus, right caudate nucleus and right hippocampus, thalamic local shape volume being associated with semantic fluency and attentional composite score (Chung et al., 2017). However, multivariable logistic regression revealed no significant differences between converters and non-converters regarding the extent of WMH or within cholinergic pathways (Hanning et al., 2019), whereas others showed that PD-MCI converters had larger WMH volume and higher hyperintensitive score compared with non-converters (Sunwoo et al., 2014).

PD-NC and PD-MCI patients who progress to PDD at follow-up are classified as “PDD converters,” who, at baseline have shown lower GMV in prefrontal and left insular cortex extending to posterior cortical areas, bilateral striatum, and the amygdala/hippocampus structural covariance network (Chung et al., 2019; Filippi et al., 2020; Zhu et al., 2022). These and other changes in PDD converters have been reviewed recently (Jellinger, 2024) and are not the subject of the present paper.

Structural brain atrophy may not be associated with any cognitive domain, with the exception of visuospatial measures in primary sensory and motor cortices, whereas functional connectivity (FC) is usually associated with attention, EF, language, learning, memory, visuospatial, and global cognition in multiple brain areas (Wylie et al., 2023).

The high-level cognitive processes are supported by intrinsic brain networks, as evidenced by resting-state functional MRI (rs-fMRI), and brain connectivity markers can predict MCI (Lin et al., 2021).

Eight major hierarchically organized resting state networks (RSN) serving higher-level cognitive functions have been identified, namely the visual (VN), sensori-motor (SMN), dorsal attention (DAN), ventral attention (VAN), limbic (LN), fronto-parietal (FPN), salience (SAN), and default mode (DMN) networks (Yeo et al., 2011; Seitzman et al., 2019), plus the executive (EN) and the subcortical (striatal; SCN) networks. Four of these RSNs (SMN, DAN, VAN, and FPN) have specifically been implicated in the pathological dysfunctions present in PD-MCI patients (see Figure 1).

The DMN, which is believed to serve an important role in various cognitive functions, includes the medial parietal, bilateral inferior-lateral parietal and ventromedial-frontal cortex (Smith et al., 2009). In PD, changes in its connectivity have been reported (Tessitore et al., 2012; Disbrow et al., 2014; Yao et al., 2014). Alterations in regions related to the DMN in PD (Tahmasian et al., 2017) support findings which indicate its involvement in cognitive decline in PD (Wolters et al., 2019). The FPN is a critical component in EFs and working memory (Petersen and Posner, 2012) and related to dopamine depletion in PD (Lang et al., 2019), while attentional orienting is driven by both the VAN and DAN (Peraza et al., 2017; Bezdicek et al., 2018), which has been associated with PD-MCI (Lebedev et al., 2014; Caminiti et al., 2015). They are associated with the cholinergic neurons in the basal forebrain, and the cholinergic neurotransmitter loss has been linked to attention deficits in this system (Ballinger et al., 2016). The SMN serves primary motor functions, but it has also been associated to both impaired sensory integration for motor function (Chen et al., 2021) and the fronto-executive dysfunction cognitive profile in PD, mainly impacting the FPN. Thus, it plays an important role in verbal short-term memory, coordinating with the fronto-temporal areas (Buchsbaum and D'Esposito, 2019). Furthermore, the SMN cognitive role arises from its disruption in association with CI in PD (Agosta et al., 2010; Chen et al., 2020). The SAN, thought to be involved in maintaining vigilance and arousal, is composed of inferior anterior insula and anterior cingulate cortex (Seitzman et al., 2019). It is a key neuronal substrate of CI in PD (Yeager et al., 2023).

Three networks, the DMN, FPN, and SAN have been linked to EF, with the SAN playing a role in modulating DMN and FPN activity (Seeley et al., 2007; Sridharan et al., 2008; Bressler and Menon, 2010). The SAN integrating responses to salient stimuli (Menon and Uddin, 2010), its control over the DMN and FPN is dysregulated in PD and aging (Putcha et al., 2016; Chand et al., 2017). The basal ganglia network (BGN) is also of significance to PD, because dopaminergic deficits in the basal ganglia can profoundly alter functional brain networks (Obeso et al., 2000; Shafiei et al., 2019; Shima et al., 2023). The relationship of SAN dysfunction to other brain networks and CI are unclear (Badea et al., 2017), as is BGN connectivity with essential cortical networks (DMN, SAN, and FPN; Yeager et al., 2023).

New techniques for quantification of structural changes across the entire network are emerging that are likely to show sensitivity for changes in PD-MCI (Nigro et al., 2016). Compared to HCs, PD-MCI patients displayed reduced between-network connectivity of large functional networks related to cognition, which increased with time and MCI status, reflecting compensatory efforts (Klobušiaková et al., 2019). Brain connectivity analyses revealed network connectivity changes in both PD-NC and PD-MCI groups compared to HCs. It should be emphasized that FC changes within the most relevant neurocognitive networks were already detectable in early drug-naive PD patients even in the absence of compensatory mechanisms (De Micco et al., 2023). Compared to HCs, PD-MCI patients had a large BGN and FPN with decreased FA in the right hemisphere and a subnetwork with increased MD involving similar regions bilaterally. The PD-MCI group showed a significant reduction in FC between the DMN and the precentral gyrus, middle temporal cortex insula, middle frontal gyrus and anterior inferior parietal lobule and a significantly decreased FC in the middle frontal and temporal gyri. Furthermore, PD-MCI patients had a network with decreased FA, including basal ganglia and fronto-temporo-parietal regions bilaterally (Galantucci et al., 2017).

Disruption of structural connections between brain areas forming a network contributed to differentiate between PD with and without MCI (Galantucci et al., 2017). The PD-MCI group had significantly decreased FC within the DMN, mainly between the hippocampal formation and inferior frontal gyrus, between the posterior cingulate gyrus and posterior inferior parietal lobule, and between anterior temporal lobule and inferior frontal gyrus. The decreased FC between anterior temporal lobe and middle temporal gyrus was positively correlated with attention/working performance, and the reduced FC between hippocampus and inferior frontal gyrus with memory function. These findings indicated that an altered DMN connectivity characterizes PD-MCI (Hou et al., 2016). Loss of GM was observed in the DMN (bilateral precuneus) without a corresponding disruption of functional properties, whereas these appeared in the SAN. There was a correlation between visuospatial scores and right supramarginal gyrus node degree. These findings highlight the loss of FC and topology without structural damage in the SAN regions in PD-MCI. Functional disruption in the absence of GM atrophy suggests that the SAN is a key vulnerable system at the onset of PD-MCI (Aracil-Bolaños et al., 2019). Comparison between PD-NC and HCs further showed a selective decline in interconnectedness between bilateral lentiform nuclei, and between PD-NC and MCI in the bilateral superior parietal lobules and precuneus. The connectivity changes were localized in the hubs of the posterior attention network. Aligned with the predominant attention deficit, this seemed to be a hallmark of PD-MCI (Bezdicek et al., 2018).

Both PD-NC and PD-MCI patients compared with HCs showed decreased DMN connectivity, while only PD-MCI ones showed decreased FC of bilateral prefrontal cortex within the FPN. In both early PD-NC and MCI PD, the topological organization of essential GM networks—SMN, DMN, SAN, and FPN—were disrupted, as well as the connections between DMN and cerebellar network, which were greater and more extended in PD-MCI (Suo et al., 2021): a significant difference between PD-MCI vs. PD-NC and HCs was in the FC-trait comprising SMN, DAN, ventral attention VAN and FPN networks, which was associated with working memory, and memory. Intra-network changes were found in resting state networks related to attention, EF and motor control. Interaction between the SMN, and DAN, VAN and FPN networks reflects the intertwined decline in motor and cognitive abilities in PD-MCI. This suggests that the memory impairment observed in PD-MCI is associated with reduced FC within the SMN and between SMN and attention networks (Delgado-Alvarado et al., 2023).

The decreased prefrontal cortex connectivity correlated with cognitive parameters but not with other clinical parameters, suggesting that altered DMN connectivity characterizes PD patients regardless of cognitive status, whereas functional disconnection could be associated with PD-MCI even in the absence of detectable structural changes (Amboni et al., 2015). Decreased cortical thickness in left superior temporal and fusiform, right insula and fusiform areas in PD-MCI resulted in decreased FC between these affected regions contributing to cognitive decline (Zhu et al., 2022). PD-MCI showed 17 WM circuits with reduced connectivity compared to HCs, mainly involving temporal/occipital regions. Reduced structural connectivity in fronto-striatal and cortico-cortical connections was associated with PD-MCI (Inguanzo et al., 2021). It further induced interrupted, slow evolution of subnetworks, predominantly linking temporal-parietal-occipital lobes, in functional brain network dynamics in early PD, and the dynamic expression characteristics of these subnetworks also reflected the degree of CI in early PD (Chu et al., 2023).

FC analysis performed with cholinergic forebrain nuclei, for the right NBM/Ch4 exhibitd lower FC in the right middle cingulate and paracingulate gyri, middle frontal, left inferior parietal and superior frontal gyri compared to HCs, and significantly lower FC in left putamen, middle frontal gyrus, cingulate and paracingulate gyri compared to PD-NC. Increased FC was seen in right calcarine fissure, surrounding cortex and cerebellum. Altered FC involved the cortical regions of NBM/Ch4. For the nucleus of the diagonal band CH1-3, FC values in the right middle cingulate and paracingulate gyri were reduced. These changes indicate functional alterations within the cholinergic system (Zhang P. et al., 2023).

PD-MCI was associated with reduced FC of the mediodorsal thalamus and para- and posterior cingulate cortex, while its connection to posterior cingulate cortex was increased (Owens-Walton et al., 2021): further decreased FC between the striatal networks was partly due to atrophy within the SAN, as were the connections between right caudate nucleus an anterior circulate cortex, precuneus and left supramarginal gyrus, while FC to left hippocampus and right cerebellar hemisphere was increased (Lang et al., 2020). PD-MCI further demonstrated poor olfactory functioning and abnormalities detected by diffusion tensor imaging in the anterior olfactory structures, relative to PD-NC individuals. Olfactory defect and microstructural changes in the anterior olfactory structure may be an additional biomarker of PD-MCI (Stewart et al., 2023).

In conclusion, both early and PD-MCI are characterized by reduced FC in the DMN and SAN, the inferior frontal and anterior temporal cortex, the left and right FPN bilateral superior medial frontal cortices, with increased FC in SAN and FPN, but breakdown of the connection between mediodorsal thalamus and the cingulate cortex, as well as the striatal network and its cortical connections, indicating a dysfunction of multiple attentional cognitive control and other essential networks.

18F-Fluorodeoxyglucose PET (18F-FDG PET) studies in PD-MCI patients revealed reduced metabolism in the frontal lobe and to a lesser degree in parietal and occipital areas compared to HCs. Mini-mental state examination (MMSE) correlated positively with metabolism in several lobes, EF with metabolism in the parieto-occipito-temporal junction and frontal lobe, visuospatial function with parieto-occipital and language with frontal metabolism (Garcia-Garcia et al., 2012). In PD-MCI patients hypometabolism exceeded GM atrophy in angular gyrus, orbital, anterior frontal and occipital lobe, indicating a gradient of severity of cortical changes associated with the development of CI (González-Redondo et al., 2014). Other PET studies in PD-MCI showed reduced metabolism in posterior cortical regions, particularly in parietal and occipital cortex, which were not affected in PD-NC (Homenko et al., 2017; Schrag et al., 2017). In comparison to both PD-NC and HCs, PD patients with MCI exhibited hypoperfusion in the parietal memory network and decreased precuneus FC in the right striatum, which was positively associated with memory dysfunction (Jia et al., 2019). PD-MCI patients had reduced regional cerebral blood flow (CBF) in the left precentral and middle cingulate gyrus, right middle frontal gyrus and bilateral putamen compared to the aMCI group. Correlations to EF were found in the PD-MCI group, and correlations to memory peformance in the aMCI group (Shang et al., 2021). After 2-year follow-up, PD-MCI patients showed significantly reduced CBF in multiple prefrontal regions. More importantly, converters to PD-MCI showed more significant CBF reduction in the left lateral orbitofrontal cortex than non-converters. These findings suggest that longitudinal CBF reduction in the prefrontal cortex might impact cognitive function (especially EF) at early stages of PD (Wang J. et al., 2022). Brain metabolism related to MCI and phenoconversion in patients with isolated REM sleep behavior disorder (iRBD), a risk facor for subsequent PD, was reduced in the inferior parietal lobule, lateral and medial occipital and middle and inferior temporal cortex bilaterally compared with HCs and the iRBD-NCI group (Yoon et al., 2022). A cognitively stable PD-NC group had frontal predominant hypometabolism, while PDD converters showed parieto-occipital hypometabolism at baseline regardless of whether a patient's initial cognitive status was PD-NC or PD-MCI (Baba et al., 2017). In PD-MCI, impaired self-awareness of cognitive deficits was related to decreased metabolism in the right superior temporal lobe, insula and midcingulate cortex, but not to cortical thickness in these regions (Maier et al., 2023). Decreased metabolism in left frontal and posterior cortex was related with executive and memory dysfunction, occipital hypometabolism with visuospatial dysfunction, suggesting that its level in a specific brain region may indirectly reflect the relevant cognitive dysfunction (Zhihui et al., 2023). MCI subgroups showed overlapped impaired regions. They had reduced cerebral blood flow in the bilateral putamen, left precentral, and middle cingulate gyrus, and right middle frontal gyrus compared to aMCI. CBF in the left precentral, left middle cingulate and right middle frontal gyrus was significantly altered (Shang et al., 2021).

PD-MCI in the posterior cingulate cortex exhibited reduced N-acetyl aspartate (NAA), total NAA, choline, glutathione, glutamate + glutamine, and total creatine (tCr) levels but both elevated myo-inositol (Ins) and Ins/tCr ratio, but reduced NAA/Ins ratio. ROC curve analysis revealed that tCr concentration could differentiate PD-MCI from controls, while PD-NC individuals with low NAA and tCr may be at risk of preclinical PD-MCI (Huang M. et al., 2023).

Earlier 18F-florbetapir PET studies in PD-NC subjects showed extremely low β-amyloid (Aβ) positivity (Mashima et al., 2017), although regional cortical Aβ deposition in the frontal cortex, precuneus and anterior or posterior cingulate gyrus inversely correlated with naming and verbal memory performance, respectively (Akhtar et al., 2017), while others found no such association (Ko et al., 2017; Melzer et al., 2019). More recent studies detected higher Aβ deposition in prefrontal and temporal cortex in PD-MCI (Ghadery et al., 2020). Aβ-positive aMCI cases had lower dopamine activities in the left striatum, suggesting changes related to AD pathology (Oh et al., 2021); others found significant correlation between plasma Aβ-42 level and FA in multiple regions of the left brain, as well as between plasma tau levels and the average thickness of the posterior cingulate gyrus (Huang C. C. et al., 2023). The finding that Aβ deposition was located mainly within the DMN suggested disturbances in Aβ topological organization characterized by abnormal network integration/segregation and the spreading pattern of Aβ between brain modules in these patients (Kim et al., 2019). Other studies suggested that regional Aβ deposition alone has a moderate effect on predicting future cognitive decline in PD and that the patchwork effect of Aβ deposition on cognitive ability may separate CI from cognitive sparing in PD (Mihaescu et al., 2022). Cerebral Aβ worsened executive functions but not the global cognitive abilities and was not associated with middle-temporal cortex atrophy, suggesting that Aβ alone may not be the main pathogenetic determinant of cognitive deterioration in PD-MCI, but would rather aggravate deficits in domains vulnerable to PD primary pathology (Garon et al., 2022). 18F-flortaucipir PET, detecting tau deposition in brain tissue, showed no or minimal tau binding in PD-NC and minimal in PD-MCI as compared to low/moderate in PD (Bohnen et al., 2017).

While the heterogeneous pathology of PDD and MCI in other dementing disorders is well-documented (Markesbery, 2010; Halliday et al., 2014; Jellinger, 2022, 2024), little is known about the specific neuropathology of PD-MCI, since there are only few detailed studies in small cohorts. Among 698 autopsy-proven cases of PD, 16 met the clinical criteria of PD-MCI (Adler and Beach, 2010; Jellinger, 2013). They included 7 naMCI cases (mean age at death 80 years, mean duration of disease 9.7 years), 50% brainstem predominant, 31% brainstem-limbic and 19% neonatal LB stages, Braak neuritic stages 0–4 (mean 2.1), two with few neuritic plaques and 4 with Aβ plaques, 8 aMCI cases (mean age 80 years, mean disease duration 15.1 years), brainstem predominant 4, brainstem-limbic 3 and neocortical stage one. Braak NFT stage I-IV (mean 2.7), 4 each with moderate neuritic plaques and with many Aβ plaques. One single aMCI MuD case (aged 75, 11 years disease duration), brainstem type, neuritic Braak stage II, with few neuritic and many Aβ plaques. Co-pathologies included mild cerebral amyloid angiopathy (CAA) in the aMCI MuD case, mild lacunar state in basal ganglia in four cases and old brain infarct in one. Braak NFT stage in aMCI cases was higher than in naMCI ones, mild to moderate neuritic plaques were present in 43% (no differences among MCI subtypes), mild CAA in 11%, lacunar state in 25% (similar in all MCI subtypes), and old cerebral infarcts in 12% (one aMCI and naMCI case each).

Another autopsy study of 159 PD cases included 25 MCI ones (56% aMCI and 44% naMCI) with no significant differences in age, gender and disease duration. All naMCI cases were brainstem-limbic LB stage 3, in the aMCI group 22% were LB neocortical stage 4. Concomitant tau pathology was present in nine cases, both aging-related tau astrogliopathy (ARTAG) and argyrophilic grain disease (a 4R-tauopathy) were present in five cases, while two aMCI cases met the pathological criteria of progressive supranuclear palsy. No differences were found in neuritic plaque stage, total Aβ and CAA score, WM rarefication, cerebral infarct volume and APOE carrier frequency. Slight differences between MCI subtypes were higher Braak NFT stage in aMCI cases and mild increase in LB pathology in naMCI-PD cases (Knox et al., 2020).

In conclusion, the few available neuropathological data gave a heterogeneous picture of PD-MCI with a combination of various types and degrees of relevant changes and co-pathologies that were comparable with MCI in other diseases (Markesbery, 2010; Takao, 2012). Therefore, the morphological basis of the different subtypes of PD-MCI and the impact of frequent co-pathologies warrant further elucidation.

Biomarkers for CI in PD could aid in both diagnostic and prognostic evaluation and in the development of new cognitive enhancing treatments. Promising biomarkers for MCI in the preclinical stage of PD are: (1) description of the cognitive profile in several subdomains applying the MDS-TF (level II) criteria; (2) 123I-FP-CIT SPECT examining degeneration of the dopaminergic nigrostriatal pathway and/or DAT transporter and/or 18F-FDOPA PET examination; (3) FDG-PET investigating brain metabolic signature; (4) EEG markers or calculating delta/theta rhythm frequency, analyzing peak frequency, etc.; (5) assessment of plasma and/or CSF levels of αSyn and/or using of seeding amplification assays of αSyn in CSF and/or plasma (quaking induced conversion assay); (6) assessment of CSF and/or plasma Aβ-40 and−42, total (t)-tau and phosphorylated (p) tau 181; (7) measurement of CSF and/or plasma neurofilament light-chain (NfL) and glial fibrillary acidic protein (GFAP) levels; (8) assessment of serum glial cell line-derived neurotrophic factor (GDNF); (9) assessment of levels of CD8+, TNF-α, IL-6, Treg, and other inflammatory-related markers; (10) assessment of αSyn, p-tau and Aβ-42 in saliva; (11) MRI findings detecting local brain atrophy using VBM and other technologies, WMH burden and functional brain network changes; (12) assessment of cortical acetylcholinesterase activity by both PET and MRI imaging; (13) combined MRI imaging and EEG examination detecting widespread structural and EEG abnormalities; (14) detection of olfactory deficits and microstructural changes in the anterior olfactory structure. Combinations of biomarkers could be valuable for the individualized diagnosis of MCI as indicator for cognitive changes in earliest or prodromal stages of PD.

LBs in cerebral cortex are present in practically all cases of sporadic PD. Therefore, the distribution of Lewy body pathology (LBP) is possibly one route or pattern of αSyn spreading in early disease stages. LBP involves the brain, extending from olfactory bulb via brainstem, amygdala to the neocortex (Braak et al., 2004). LBP may also start in the enteric system ascending via the vagus nerve to the brain (Borghammer et al., 2021), suggesting that either αSyn aggregation begins in the gut and spreads in a prion-like fashion to the brain or systemic inflammatory processes driven by gastrointestinal dysfunction contribute to the pathogenesis of PD before the development of overt clinical symptoms (Ryman et al., 2023). Immune and inflammatory-related changes may represent another important factor in the pathogenesis of PD, inflammatory markers being associated with cognitive dysfunction in PD (Zhao et al., 2024). However, there is a clear morphological heterogeneity in PD-MCI, similar to that in MCI cases without PD (Markesbery, 2010). It includes a variable combination of LBP and AD-related changes (ADNC), both of which with much lower severity and extent than in fully developed PDD (Jellinger, 2023).

A recent review confirmed the Braak hypothesis of LBP staging from molecular, cell biological (cell-to-cell spread) and organ-level (region-to-region spread) of αSyn pathology (Carceles-Cordon et al., 2023). However, LBP does not always follow the same bottom-up progression from olfactory bulb via brainstem to the cortex, as there may be a multifocal distribution pattern of αSyn pathology (Uchihara and Giasson, 2016). It may involve the NBM, amygdala and locus ceruleus in the prodromal phase of the body-first type of PD (Borghammer et al., 2021). Recent studies provided insight into dysfunctions of dopaminergic, cholinergic and noradrenergic systems that are involved in PD-CI (Ye, 2022). Reduced presynaptic dopamine uptake in the striatum resulted in reduced prefrontal and parietal metabolism (Ekman et al., 2012) and increased signal variability in the SAN, which in turn caused decreased corticostriatal connectivity (Shafiei et al., 2019; Shima et al., 2023). PD-MCI patients compared to PD-NC showed a more widespread degeneration of dopaminergic terminals in the caudate nucleus (Sasikumar and Strafella, 2020), while degeneration of the medial substantia nigra caused dysfunction of striato-frontal and mesocortico-limbic loss (Martínez-Horta and Kulisevsky, 2011). Early cognitive deficits in PD were closely related to the degeneration of the cholinergic forebrain, early degeneration of the NBM and the nucleus of the vertical limb of the diagonal band of Broca, that precede CI in PD (Liu et al., 2018; Ray et al., 2018; Schulz et al., 2018). PD-MCI converters had significantly greater MD of both NBM tracts compared to PD-NC, which was associated with cognitive outcome (psychomotor speed, working memory, delayed recall, and visuospatial function; Crockett et al., 2023). Degeneration of the cholinergic posterior basal forebrain in PD was accompanied by functional cortical changes in acetylcholinesterase activity that was associated with multi-domain cognitive deficits in PD without dementia (Schumacher et al., 2023).

Modern neuroimaging studies documented dysfunction of multiple cortical and subcortical neuronal systems with disruption of essential interconnected brain networks which result from the presence and extension of both LBP and co-morbid ADNC into the limbic and higher cortical association areas. Common mechanisms of PD-MCI and PDD involve not only the aggregation and spreading of αSyn/LBs but also of Aβ and tau in cortical and limbic regions (Masliah et al., 2001; Esteves and Cardoso, 2020). The association of cortical αSyn pathology, Aβ and tau load suggests an interaction of these pathological proteins in the pathogenesis of PD and other neurodegenerative disorders (Clinton et al., 2010; Compta et al., 2011; Jellinger, 2012; Guo et al., 2013; Miller et al., 2022). A potential mechanism for these synergistic effect may be phosphorylation of tau protein induced by αSyn as a posttranslational modification and vice versa (Giasson et al., 2003; Walker et al., 2015; Basheer et al., 2023). Post-mortem studies have confirmed the concomitant protein action, such as the co-aggregation of αSyn, Aβ and tau in PD brains with and without CI (Han and He, 2023; Noguchi-Shinohara and Ono, 2023). It should emphasized that, in general, one of these mechanisms alone may not explain the development of CI in PD, whereas at least two or more molecular mechanisms may concur and contribute to cognitive decline in PD (Irwin et al., 2013; Jellinger, 2022). Recent autopsy studies revealed decreased synaptic density in temporal, cingulate and insular cortices in PD and in the entorhinal and parahippocampal region in PDD, which was associated with higher NfL immunoreactivity, LB density and higher CI scores. In additon, synaptic loss was linked to axonal loss and αSyn burden. These results indicate a relation between synaptic loss in specific brain regions, αSyn burden and CI in PD (Frigerio et al., 2024). Other studies suggest that both αSyn and tau pathologies and impaired insulin signaling underlie CI in PD (Blommer et al., 2023). Human αSyn overexpression in the mesencephalon leads to its increase in the hippocampus inducing altered synaptic transmission and plasticity, with deceased expression of glutamate receptors. This may cause involvement of major neuronal networks leading to memory impairment in PD (Iemolo et al., 2023).

A different pathogenetic model reflecting region-specific changes in gene expression in αSyn inducing a model of early PD appears of interest. In rats, after infusion of human αSyn oligomers (H-αSynOs) into the substantia nigra, the transcription profile in regions involved with MCI were examined. Among more than 17,000 genes expressed in the hippocampus and anterior cingulate cortex, a few were differently expressed (up- or down-regulated) and related with immune functions with decreased expression of CD68 within microglia cells. In contrast, the most significantly enriched terms in the hippocampus were involved in mitochondrial homeostasis, potassium voltage-gated channel, cytoskeleton, and fiber organization, suggesting that the gene expression in the neuronal population was mostly affected in this region in early disease stages. These results show that H-αSynOs trigger a region-specific dysregulation of gene expression in anterior cingulate cortex and hippocampus as pathological substrate of MCI in early PD (Manchinu et al., 2024). The applicability of this interesting rat model to pathology is a matter of further discussion.

In conclusion, cognitive involvement in PD-MCI is likely to arise from a combination of mechanisms causing complex and multilocal cerebral lesions that induce dysfunction of both whole-brain functional networks and dysfunction of several neuromodulator systems, especially loss of cholinergic function. The interplay of pathological accumulation of αSyn as well as tau and Aβ is essential for neurodegeneration inducing these complex pathogenic mechanisms associated with regional synaptic loss, which are responsible for cognitive decline in PD, yet their causative interplay needs further elucidation.

A range of dopaminergic or psychotropic medications, psychotherapeutic techniques, stimulation therapies, cognitive training and rehabilitation methods, and other non-pharmacological treatments have been studied and are used for managing MCI in PD, but appropriate management of CI (and other neuropsychiatric symptoms) is critical for comprehensive PD care (Weintraub et al., 2022). An earlier study suggested that L-dopa might prevent a decline in cognitive function in PD (Ikeda et al., 2017); another one hypothesized that a combined stimulation of both dopamine receptor families with dopamine agonists (rotigotine, carbergoline), L-dopa and pergolide may preserve cognitive function, whereas no differences were found in cognitive function between the basal state and treatment with dopamine agonists alone (Brusa et al., 2013). No significant difference in anticholinergic burden was found between PD-MCI and PD-NC (Sumbul-Sekerci et al., 2022), while a recent review of nine articles stated that there is evidence that rivastigmine is beneficial for rapid eye movement sleep behavior disorder (RBD) and apathy in PD without dementia and may reduce falls, which may be due to improved attention. However, the outcomes of the reviewed studies were heterogenous (Reilly et al., 2024). While anti-hypertensive medications were positively correlated with PD-MCI, anticholinergic drugs burden does not appear to modulate MCI risk in PD (Cicero et al., 2022b). However, MCI in PD patients adversely affects medication management, although cognitive areas predicting success in medication management performance are language, event-based prospective memory and visuospatial functions (Sumbul-Sekerci et al., 2023).

A variety of ketogenic therapies were utilized in the MCI groups including a ketogenic diet, low-carbohydrate or Mediterranean diet with coconut oil supplementation, a diet with a ketogenic medium chain triglyceride supplement or other ketogenic compounds. They showed statistically significant improvements in some, although not all, of cognitive measures, although only after 6 months of adherence that was problematic in most of these studies (Price and Ruppar, 2023). A systemic review of therapeutic ketogenic trials gave evidence for cognitive improvement in individuals with MCI, but the number of studies in PD is very small and further research is requested to optimize the utilization of ketogenic interventions in clinical contexts (Bohnen et al., 2023).

Cognitive rehabilitation has been found to improve specific cognitive deficits in PD, but not all cognitive domains may benefit from this method that should be tailored to patient's specific impairments (Giustiniani et al., 2022a,b). A pilot study on the effects of a new integrated and multidisciplinary cognition program based on mindfulness and reminiscence therapy showed a significant improvement in MoCA memory score and could be effective in PD-MCI patients (Reitano et al., 2023). Cognitive training can increase physical activity possibly due to effects on executive function, but this also needs further investigation in larger samples (Bode et al., 2023).

In conclusion, the effects of medical intervention in PD patients with MCI are controversial, although among anticholinergic drugs, rivastigmine may improve attention, and some cognitive improvement was reported for ketogenic diets and for both cognitive rehabilitation and mindfulness therapy. For most of these methods, the number of trials is too small and further research for their optimization and validation are warranted.

PD, a common and heterogeneous neurodegenerative disease, characterized by a combination of motor and non-motor symptoms, is frequently preceded by a period of MCI, affecting multiple cognitive domains, which may or may not progress to dementia or may revert to normal cognition. The clinical manifestations, molecular/biochemical and morphological basis of PD-MCI are heterogeneous, and modern neuroimaging studies revealed widespread changes in cerebral GM and WM, involving multiple brain areas and causing disruption of many critical neuronal networks involved in cognitive, attention and memory functions as well as various neuromodulator systems, lesions which often antedate structural changes. Given the inherent heterogeneity in the clinical presentation, neuropsychology, functional morphology and neuropathology of PD-MCI, diagnostic criteria regarding definition and classification have been established, and modern neuroimaging studies have given insight into the underlying neurobiological mechanism including early metabolic changes. Artificial intelligence on FDG-PET images has identified MCI patients not only in PD but also in other neurodegenerative diseases with 75% sensitivity and 84% specificity (Prats-Climent et al., 2022). The pathogenesis, among others, depends on the extent and severity of LBP and ADNC co-pathologies, mainly involving limbic and subcortical brain areas with less severe extension to the neocortex. A deeper understanding of the essential molecular/biochemical and neurotransmitter changes and in particular of the responsible brain network disorders is required to better understand the relationship of the different types of PD-MCI and their relations with other clinical symptoms. The prospective assessment and validation of MCI and a deeper understanding of the interaction of multiple genetic factors will be achieved by modern fluid biomarkers (e.g., reduced αSyn and Aβ levels, increased p-tau 128, GFAP, and NfL), as well as inflammatory, EEG and diffusion and structural MRI markers. These and other studies will provide a deeper understanding of the pathophysiological processes underlying the timely progress and development of cognitive decline in relation to the neurodegenerative changes in PD-MCI as a basis for the development of effective disease-modifying therapies and preventive measures to slow or halt progression of cognitive impairment in this debilitating disease.

KJ: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Validation, Visualization, Writing—original draft, Writing—review & editing.