Associations of Alzheimer's-related plasma biomarkers with cognitive decline in Parkinson's disease

We recruited 70 consecutive patients with PD admitted to Fujita Health University Hospital between May 2020 and September 2021. All patients with PD fulfilled the Movement Disorder Society (MDS) clinical diagnostic criteria [18]. Patients with PD were further subcategorized into nondemented Parkinson’s disease (PDND) and Parkinson’s disease dementia (PDD) groups [19], with the latter diagnosed based on the MDS Task Force algorithm [20]. In the PD group, 21 patients (30.0%) were diagnosed with PDD, while the remaining 49 patients were classified as having PDND. We also enrolled 38 age- and sex-matched HCs from our ongoing aging cohort study at Fujita Health University, Japan. HCs were included based on the following criteria: (1) cognitively normal with Mini-Mental State Examination (MMSE) scores greater than 25 [21] and an Addenbrooke’s Cognitive Examination-Revised (ACE-R) total score greater than 88 [22] without a history of neurological or psychiatric disorders and (2) no observable anatomical abnormality in the brain according to magnetic resonance imaging.

This study was approved by the ethics committee of Fujita Health University Hospital, and all participants provided written informed consent before participation, as well as opt-out consent.

Motor and nonmotor symptoms related to PD were assessed using the Japanese version of the Movement Disorder Society’s Unified PD Rating Scale (MDS-UPDRS) [23]. Cognitive performance was evaluated using the Frontal Assessment Battery (FAB) [24], ACE-R, MMSE, and the Japanese version of the Montreal Cognitive Assessment (MoCA-J) [25]. The MoCA was evaluated using the total score and six subscores consisting of attention, orientation, executive, memory, language, and visuospatial domains, which were proposed in the original MoCA manuscript [25]. The ACE-R was evaluated using the total score and five subscores comprising orientation and attention, memory, verbal fluency, language, and visuospatial abilities. We also assessed the participants using the Parkinson’s Disease Questionnaire-39 Summary Index, Geriatric Depression Scale-15, Odor Stick Identification Test for the Japanese (OSIT-J) score, Japanese version of the REM Sleep Behavior Disorder Screening Questionnaire, Scales for Outcomes in Parkinson’s Disease-Autonomic, Epworth Sleepiness Scale, and Japanese version of the Questionnaire for Impulsive − Compulsive Disorders in Parkinson’s Disease. The levodopa equivalent daily dose was calculated according to established formulae [26] with additional consideration of opicapone and safinamide intake [27]. We performed all neurological evaluations of patients with PD during the ‘on’ condition.

Table 1 illustrates the clinical characteristics of individuals in the PD and HC groups. Patients with PD exhibited significantly lower education levels (p = 0.0146), higher depression scores (Geriatric Depression Scale-15: p < 0.0001), and lower scores on global cognitive scales (MMSE, ACE-R total score, and MoCA-J total score: p < 0.0001) than did HCs but did not differ in age at examination and sex. Supplementary Table 1 summarizes the clinical characteristics of the individuals in the HC, PDND, and PDD groups. Notably, patients with PDD displayed a significantly older age at examination (p = 0.0104) and onset (p = 0.0127), a lower education level (p = 0.0121), a higher Hoehn and Yahr (HY) scale (p = 0.0372), more severe hyposmia (OSIT-J score: p = 0.0003), and lower scores on global cognitive scales (MMSE, ACE-R total score, and MoCA-J total score: p < 0.0001) than did patients with PDND.

Figure 1a and b demonstrates that the GFAP and NfL levels were significantly higher in the PD group than in the HC group (PD group: 144 ± 65.6 pg/mL, HC group: 115 ± 39.1 pg/mL, p = 0.0363; and PD group: 39.8 ± 32.5 pg/mL, HC group: 13.9 ± 4.45 pg/mL, p < 0.0001; respectively). Conversely, there was no significant difference in the p-tau181 level between the two groups (PD group: 2.15 ± 1.33 pg/mL, HC group: 1.62 ± 0.684 pg/mL, p = 0.0989) (Fig. 1c). With regard to Aβ, patients with PD exhibited significantly higher Aβ42/40 levels than did HCs (PD group: 0.0709 ± 0.0128, HC group: 0.0646 ± 0.0139, p = 0.0011) (Fig. 1d), whereas the analyses of Aβ40 and Aβ42 separately showed no significant difference between the PD and HC groups (PD group: 96.0 ± 20.0 pg/mL, HC group: 100.0 ± 15.8 pg/mL, p = 0.2429; and PD group: 6.74 ± 1.67 pg/mL, HC group: 6.43 ± 1.49 nmol/h/mL, p = 0.3940; respectively) (Fig. 1e, f).

Table 2 summarizes the relationships between the six plasma biomarkers and clinical indices in the PD group. Plasma GFAP, NfL, p-tau181, and Aβ40 levels were significantly positively correlated with age at examination and age at onset. Significant positive correlations were found between GFAP and MDS-UPDRS I scores, NfL and MDS-UPDRS II scores, and GFAP, NfL, p-tau181, or Aβ40 and the HY scale. GFAP also showed a significant negative correlation with OSIT-J scores. Furthermore, NfL and especially GFAP were significantly negatively correlated with MMSE, ACE-R, MoCA-J, and FAB scores (Fig. 2a–h). However, Aβ40 was weakly correlated with MMSE and ACE-R scores, Aβ42/40 was only weakly correlated with MMSE scores, and p-tau181 was weakly correlated with ACE-R and FAB scores. Notably, GFAP was negatively correlated with all ACE-R and MoCA-J subscores; NfL was negatively correlated with all but the language subscore of the MoCA-J, while p-tau181, Aβ42/40, Aβ40, and Aβ42 showed significant correlations with only some subscores (Table 3). Conversely, in the HC group, these six plasma biomarkers showed no significant correlation with MMSE, ACE-R, or MoCA-J scores.

Figure 1g–l and Supplementary Table 1 show the multiple comparison results of the six plasma biomarkers among the HC, PDND, and PDD groups. Steel–Dwass post hoc comparisons showed that patients with PDD had higher GFAP levels than did HCs (p < 0.0001) and patients with PDND (p = 0.0008) (Fig. 1g). The NfL levels significantly differentiated PDD patients from patients with PDND (p = 0.0291) and differentiated both patients with PDND (p < 0.0001) and those with PDD (p < 0.0001) from HCs (Fig. 1h). Regarding p-tau181 levels, there was a significant difference between HCs and patients with PDD but not between HCs and patients with PDND (Fig. 1i). Although no significant differences were observed in multiple comparisons of Aβ40 and Aβ42 (Fig. 1k, l), patients with PDND presented significantly higher Aβ42/40 levels than those of HCs (p = 0.0001) and PDD patients (p = 0.0195) (Fig. 1j). There were no significant differences between patients with PDD and HCs (Fig. 1j). Regarding the ROC analysis conducted on the six plasma biomarkers across various groups (Fig. 3 and Table 4), NfL and Aβ42/40 effectively distinguished the HC and PDND groups (AUC values: NfL, 0.8457; Aβ42/40, 0.7597), whereas GFAP and NfL exhibited excellent discriminatory potential in distinguishing the HC and PDD groups (AUC values: GFAP, 0.8396; NfL, 0.9148; GFAP*NfL, 0.9236). Notably, GFAP and Aβ42/40 demonstrated significant discriminatory capacity in differentiating between the PDND and PDD groups (AUC values: GFAP, 0.7765; Aβ42/40, 0.7046).

Spearman’s rank correlation coefficients, along with their respective p values, are summarized in a correlation matrix (Supplementary Fig. 1). Within the PD group, a substantial number of correlations was observed between the interrelationships among the six biomarkers. Conversely, in the HC group, only the relationships between Aβ42 and Aβ40 levels, Aβ42 and Aβ42/40 levels, and NfL and p-tau181 levels showed significant positive correlations.

Table 5 presents the results of the generalized linear model analysis of the effects of plasma biomarkers and possible confounders on the ACE-R and MoCA-J scores. Education, GFAP levels, and NfL levels were significantly associated with the ACE-R scores, while GFAP and NfL levels showed a tendency to correlate with the MoCA-J scores, but not significantly.

Plasma GFAP serves as a biomarker for astrocytic activation [28], and reactive astrogliosis is increasingly being implicated in PD pathogenesis [29]. Pathological studies of the post mortem brains of patients with Lewy body disorder have revealed that astrocytic alpha-synuclein accumulation contributes significantly to alpha-synuclein pathology [30]. Although the relationship between plasma GFAP levels and clinical scores in PD patients remains incompletely understood, recent studies have demonstrated higher plasma GFAP levels in PD patients with dementia and mild cognitive impairment than in controls [14, 31]. Moreover, these studies identified a significant negative correlation between plasma GFAP levels and MMSE scores in all participants with PD.

Neural networks that cause cognitive symptoms in patients with PDD are widely distributed and diverse, with overlapping functions that depend on primary neurotransmitters [32]. Evidence suggests that damage to one network may influence another, as neurotransmitters can modulate each other’s effects [33], although the cellular-level pathology in PDD is heterogeneous, and the effects of different genes are still being uncovered [32]. Our findings revealed that plasma GFAP levels were negatively correlated with not only the total MMSE, MoCA-J, ACE-R, and FAB scores but also with all subcategories of the MoCA-J and ACE-R. These findings suggest that astrocytic pathology may be a common feature of cognitive decline in PD irrespective of the involvement of different cognitive networks.

Plasma NfL is a component of the neuronal cytoskeleton, and increased plasma NfL has been shown to serve as a biomarker for a variety of neurodegenerative diseases [11]. Although controversy remains as to whether plasma NfL levels are elevated in patients with PD compared with those in HCs [14, 34‐36], the NfL value has been significantly associated with both motor severity and cognitive decline in PD [37‐39], suggesting its usefulness as a disease progression marker in PD. We confirmed that plasma NfL was negatively correlated with MMSE, ACE-R, MoCA-J, and FAB scores, although the correlations tended to be weaker than those of GFAP. Plasma NfL, in contrast to GFAP, was also positively correlated with the MDS-UPDRS Part II and HY scale. GFAP and NfL could be biomarkers of cognitive decline in PD, but it should be considered that NfL can also reflect motor symptoms.

Plasma p-tau181 is a useful diagnostic and prognostic biomarker of AD, correlates with cerebrospinal fluid (CSF) p-tau181, and predicts Aβ and tau positivity on PET [10]. Evidence suggests that tau is involved in the pathophysiology of PD, with tau and alpha-synuclein colocalizing in Lewy bodies [40]. Moreover, a genome-wide association study identified MAPT, the gene encoding the tau protein, as a risk factor for PD [41].

Higher plasma p-tau181 levels have been reported in patients with PD than in HCs, but no significant association has been observed between plasma p-tau181 levels and PD-related clinical indices, including the HY and cognitive scales [13, 35]. In our study, patients with PD showed a tendency toward higher plasma p-tau181 levels than those in HCs, but the difference was not significant. Only patients with PDD showed a significant increase in plasma p-tau181 levels when compared with HCs. Although the clinical significance of plasma p-tau181 in PD remains unclear, our study suggests its potential relevance in advanced cognitive decline in PD.

Previous studies have demonstrated that plasma Aβ42/40 levels decrease in individuals with amyloid PET-positive AD [9, 42]. Although the global Aβ load on PET is negatively associated with memory and language functions [43], the relationship between plasma Aβ42/40 levels and Aβ pathology in the brain has not been fully investigated in PD.

Our study found that plasma Aβ42/40 levels were significantly higher in patients with PD, especially PDND, than those in HCs, which is consistent with a previous report regarding the use of plasma biomarkers for the differential diagnosis of Parkinson syndromes [44]. The detailed mechanisms underlying the higher plasma Aβ42/40 levels in patients with PD remain unknown, but one possible explanation is that alpha-synuclein uptake may interfere with monomeric Aβ40 [45]. Another study showed decreased Aβ40 levels with increased alpha-synuclein levels in patients with PD [46]. A head-to-head comparison study of plasma biomarkers in multiple system atrophy, a synucleinopathy similar to PD, demonstrated decreased Aβ40 and increased Aβ42/40 levels, which supports this hypothesis [47]. Although plasma Aβ40 levels were not significantly decreased in patients with PD compared with those in HCs, alpha-synuclein may influence plasma Aβ levels in patients with PD.

However, multiple comparison results showed decreased plasma Aβ42/40 levels in patients with PDD compared with those in PDND patients but not with those in HCs. ROC analysis also showed that decreased plasma Aβ42/40 levels could be a supportive finding to distinguish patients with PDD from those with PDND. This could suggest that amyloid pathology is more developed in patients with PDD than in patients with PDND which is consistent with a previous meta-analysis demonstrating the involvement of amyloid pathology in the development of PDD [48]. Our study indicated plasma Aβ42/40 levels might exhibit divergent changes in patients with PD without cognitive impairment and in patients with PDD.

Because plasma Aβ40, Aβ42, and Aβ42/40 levels showed limited correlations with plasma GFAP, NfL, and p-tau 181 levels compared with those of AD patients, the dynamics of plasma Aβ42/40 in PD may differ from those of amyloid PET and CSF Aβ42/40. Further studies in larger PD cohorts with PET imaging, CSF, and blood measurements of Aβ and alpha-synuclein pathology are needed to provide more insight.

The results of this study suggest that in cognitive impairment in PD, elevated plasma GFAP and NfL levels precede the appearance of abnormalities in Aβ 42/40 and p-tau181 levels, which occur only after marked progression of dementia. Pathological analysis of post mortem brains showed an increased severity of alpha-synuclein pathology in the limbic and cortical regions of PDD patients compared with those of PD patients with normal cognition, but no change was observed in the severity of tau or Aβ pathology [49]. Autopsy studies have reported an association between plasma GFAP and AD pathology including Aβ and tau in Lewy body spectrum disorders [50]. In contrast, our results showed that Aβ42/40 levels were decreased in the PDD group compared with those in the PDND group and that p-tau181 levels were increased in the PDD group compared with those in HCs. This finding is consistent with the reports that PD with Aβ accumulation is a significant predictor of cognitive decline [51] and that approximately one-third to half of patients with PDD exhibit abnormal Aβ accumulation on PET [52, 53].

Traditionally, cognitive dysfunction in PD is thought to be preceded by frontal lobe-based deficits in working memory, executive function, and attention [54]. However, the recent investigations suggest that a wide range of domains is impaired, including executive function, memory, visuospatial function, attention, and language [55]. The pathogenesis of this clinical condition is difficult to explain by the alpha-synuclein propagation hypothesis and complications of AD pathology. The failure of numerous clinical trials targeting cognitive decline in PD to date also suggests a need to consider a new pathological hypothesis [56]. Our study showed that the etiology of cognitive decline in PD is intricate, with astrocytic lesions playing a significant role, while the impacts of alpha-synuclein, tau, and Aβ pathology may differ depending on the severity of cognitive decline. AD-related plasma biomarkers could be valuable in elucidating the underlying pathological mechanisms of cognitive decline in PD and in developing preventative measures to mitigate this condition.