Introduction
The development of dementia is a key milestone in the progression of Parkinson’s disease (PD). Almost half of patients develop PD dementia (PDD) within 10 years from diagnosis [
76], reaching over 80% at 20 years [
29]. Widespread cortical and limbic Lewy body deposition has been reported by several clinicopathological studies to be the best pathological correlate of cognitive decline in PD [
1,
30,
34,
38,
47,
51]. However, the association between cortical Lewy body pathology and PD dementia is far from clear-cut, given that approximately one-third of PD cases classified as Braak PD Stage 3 (indicative of no neocortical Lewy bodies) were found to be demented during life [
5]. Conversely, a number of PD cases exhibiting neocortical and/or limbic Lewy body pathology had no history of cognitive impairment [
8,
60]. Other studies have reported a significant role for co-existing Alzheimer’s type pathology [
10,
27,
37,
39,
41,
64] and a combination of both cortical Lewy body and Alzheimer’s-type pathologies has been suggested as a more robust correlate of PD dementia [
9,
31]. Despite extensive research on the neuropathological substrate of PDD, a consensus has yet to be reached. These conflicting results may be in part due to differences in case selection, the methodologies used, as well as the inherent heterogeneity of the disease, however, it also suggests that mechanisms other than protein aggregation may be critically contributing to cognitive decline in PD.
Neuroinflammation in the PD brain has been described in a small number of postmortem studies, as well as in vivo using [
11C]PK11195 PET imaging [
25], but clinicopathological studies assessing neuroinflammation in PD dementia cases are lacking. McGeer and colleagues were the first to report an increase in the number of HLA-DR
+ microglia in the substantia nigra of PD compared to healthy brains [
52], while a subsequent study found increased numbers of microglia in the hippocampus, transentorhinal, cingulate and temporal cortices of PD cases compared to controls [
36]. Infiltration of both helper (CD4
+) and cytotoxic (CD8
+) T lymphocytes into the parenchyma of the substantia nigra has been observed in the vicinity of neuromelanin-positive dopaminergic neurons in the PD brain [
6]. Upregulation of pro-inflammatory cytokines has also been reported in PD, including increased expression of tumour necrosis factor α (TNFα), interleukin 1β (IL-1β), and interferon γ (IFNγ) in the substantia nigra, and upregulation of interleukin 6 (IL6) and interleukin 2 (IL2) in the striatum [
33,
55‐
57]. Pro-inflammatory cytokine expression has not been explored in more widespread brain regions.
A prominent pathway regulating inflammatory responses is mediated by Toll-like receptors (TLRs) [
42]. Accumulating evidence suggests that α-synuclein may be triggering microglial activation via TLR2 and TLR4, leading to downstream secretion of pro-inflammatory mediators [
13,
22,
44,
46]. Both these receptors have been found to be upregulated at the protein level in the caudate/putamen of postmortem PD cases compared to controls [
16], with TLR4 also being elevated in the substantia nigra [
67] and TLR2 in the anterior cingulate cortex [
19].
Hence, although there is accumulating evidence suggesting that neuroinflammation is a feature of the pathology of PD, investigation of neuroinflammatory processes in extra-nigral brain regions has been limited to date, and no studies have explored associations with cognitive status during life. We therefore sought to characterize inflammatory changes across multiple brain regions in demented compared to non-demented PD cases and age-matched controls, in addition to better characterizing the anatomical pattern of misfolded protein pathology in these cases. We explored the relationship between markers of neuroinflammation and aberrant forms of α-synuclein, tau and amyloid-β, as well as the association of neuropathological findings with cognitive decline during life.
Discussion
This study provides novel insights into the neuropathological substrates of cognitive decline in PD through investigating, for the first time, the nature and distribution of neuroinflammatory change in PD cases with or without dementia and correlating this with protein pathology. We confirm previous findings that α-synuclein pathology correlates with the rate of cognitive decline in PD, whilst the levels of Alzheimer’s disease-type pathology were found to be comparable across groups. Neuroinflammatory change in PDD cases was most pronounced in the amygdala, a limbic region heavily implicated in emotion and cognition [
61]. Specifically, we observed increased microglial activation in the amygdala in PDD brains compared to controls and found evidence of CD4
+ T cell infiltration into this region in all PD dementia cases. Furthermore, microglial activation, CD4
+ T cell infiltration and α-synuclein pathology were correlated in this region, implicating an α-synuclein-driven neuroinflammatory response in the amygdala in PD dementia. We explored the expression of pro-inflammatory cytokines as well as TLR2 and TLR4 in several extra-nigrostriatal regions and observed elevated expression of TLR4 in the amygdala, frontal cortex, and substantia nigra, accompanied by elevated levels of the downstream inflammatory cytokine IL-1β. Taken together, our observations are consistent with the hypothesis that α-synuclein drives a neuroinflammatory response in PD through the activation of microglial TLR4 [
46] and suggest a contributory role for peripheral T lymphocytes.
Our results show a significantly higher burden of α-synuclein pathology in PDD compared to PDND cases across multiple brain regions. This is consistent with previous findings of increased α-synuclein burden in PD dementia. Compta et al. using both a semi-quantitative Lewy body scoring system and quantification of Lewy body density per mm
2 found a significantly elevated burden in PDD (n = 29) compared to PDND brains (n = 27), particularly in the frontal, temporal, cingulate and entorhinal cortex [
9]. A second large autopsy study in 92 PDD and 48 PDND brains using traditional scoring protocols reported similar results in the same brain regions [
38]. These findings were further validated in another large study with 55 PDD and 49 PDND cases [
64]. Previous volumetric MRI studies have also implicated the amygdala in PD dementia showing significant atrophy of this region in demented PD patients compared to healthy controls but not in cognitively intact PD patients compared to controls, thus implicating the amygdala in cognitive decline in PD [
40]. A limited number of studies have also addressed the role of abnormalities in the amygdala in relation to other non-motor symptoms of PD. On a functional level, a magnetic resonance imaging (MRI) study revealed that in the absence of structural alterations, there were abnormally high levels of activity in the amygdala of depressed PD patients compared to patients without depression and to controls. This heightened activity was found to be positively correlated with clinical scores of depression. Functional connectivity between the amygdala and fronto-parietal cortices was also found to be reduced, specifically in the patients suffering from depression [
32]. Furthermore, an early clinicopathological study showed that PD patients suffering from hallucinations had nearly double Lewy body density in the basolateral amygdala compared to patients that did not experience them [
28]. Amygdala abnormalities have also been linked to cognitive decline in Alzheimer’s disease. Similar to PD, atrophy of the amygdala was shown to be substantial in two large independent cohorts of mild Alzheimer’s disease. The magnitude of atrophy was strongly predictive of cognitive decline as shown by a robust correlation with MMSE scores [
62]. The role of amygdala dysfunction in PD dementia has not been extensively studied and based on our present findings may warrant further investigation.
Notably, we did not observe any significant differences in tau pathology between PDND and PDD cases. In agreement, semi-quantitative scoring in a large postmortem study revealed similar levels or neurofibrillary tau tangles in the temporal, mid-frontal, and parietal cortex of PDND and PDD cases [
38]. Tau Braak staging, however, has shown inconsistent results across studies; whilst Horvath et al. found the overall tau Braak stage to be significantly higher in demented compared to non-demented PD cases [
30], Ruffmann and colleagues did not find differences between the groups, with 84% of all cases having only mild tau pathology (Braak stage 0–2) [
64]. We did not observe significant differences in amyloid-β pathology between controls, PDND, and PDD cases. In contrast, previous studies have reported higher amyloid-β scores in the hippocampus, striatum, entorhinal and frontal cortex of demented compared to non-demented PD brains [
64]. The total amyloid-β plaque score, total amyloid angiopathy in the cortex [
9], and the amyloid-β Thal phases [
30,
72] have also been reported to be significantly higher in PDD compared to PDND. This discrepancy of our results with previous studies may be due to the difference in brain regions under investigation and the smaller sample size used in our study. Indeed, we observed a trend for increased amyloid-β deposition in the PDD cases compared to both PDND and controls, however this was not significant after correcting for multiple comparisons.
We hypothesized that neuroinflammation might be an additional neuropathological substrate contributing to dementia in PD. PET neuroimaging studies using [
11C]PK11195, a ligand for TSPO which is upregulated on activated microglia, have similarly suggested that microglial activation is increased in PDD cases. Edison and colleagues demonstrated increased tracer uptake in multiple brain regions in demented PD patients compared to controls, which was much more widespread than in non-demented PD patients versus controls [
20]. We have previously shown that PD patients with a higher risk of progressing to dementia have increased activation of the innate immune system, including an increase in classical (inflammatory) monocytes, and increased monocyte expression of both TLR2 and TLR4 compared to patients at low risk of dementia [
74]. Furthermore, we found that a pro-inflammatory cytokine profile in the serum in newly-diagnosed PD patients was associated with faster UPDRS-III progression and more impaired cognitive function over 3 years of follow-up [
77]. However, the contribution of neuroinflammation to PDD has not previously been explored at postmortem. Our novel data show an increase in activated HLA-DR
+ microglia in the amygdala of PDD cases. In a previous study Imamura et al. showed increased numbers of HLA-DR
+ microglia in the hippocampus, transentorhinal, cingulate and temporal cortex in a relatively small number of PD (n = 12) and control (n = 4) autopsy cases, though in this study no distinction was made between demented and non-demented PD brains [
36]. The lack of a PD versus control difference in activated microglia in these regions in our study may relate to the characteristics of the control population used. We opted to select typical elderly controls on the basis of having no neurological or cognitive symptoms during life and not on the basis of an absence of tau or amyloid-β pathology in the brain. In contrast, other authors typically select “supranormal” controls with no neurofibrillary tau tangles or amyloid-β plaques. Such controls are not representative of the normal neurologically intact aged population; indeed it has been repeatedly demonstrated that misfolded tau and amyloid-β accumulation occurs during ageing in the absence of neurodegenerative disease [
3,
11,
50,
63,
65]. Our controls had a degree of amyloid-β and tau pathology and such misfolded protein deposition may trigger low level microglial activation. This could explain the contradictory findings in our study compared to previous work [
36].
Surprisingly, we did not find a difference in HLA-DR
+ microglia count in the substantia nigra of PD cases (either PDND or PDD) compared to controls. This is in contrast to the seminal study by McGeer et al. in 1988 who first reported an increase in HLA-DR
+ microglia in this region of PD cases compared to controls [
52]. Similar findings have been reported by a subsequent study showing an increase in both amoeboid CD68
+ microglia as well as Iba1
+ microglia in the postmortem PD nigra [
15]. These inconsistencies with our findings may be partly explained by methodological differences in the identification of activated microglia as discussed below, as well as differences in the selection of control populations. [
11C]PK11195 PET neuroimaging studies have also shown conflicting data. Ouchi et al. found increased binding in the midbrain of newly diagnosed PD patients [
59], whilst Gerhard et al. did not find a difference in the substantia nigra of patients compared to controls [
25]. Additional studies are needed to ascertain the extent of microglial activation in the substantia nigra and at which stage of the disease this is more prominent.
Although we did not observe an increase in number of activated microglia in the nigra, we did observe increased infiltration of peripheral T lymphocytes in this region in PDD and PDND cases, as well as elevated IL-1β levels, providing alternative evidence of immune activation in this region. Brochard et al. have previously shown an increased number of both CD4
+ and CD8
+ T lymphocytes in the substantia nigra of PD cases compared to controls, especially in the vicinity of dopaminergic neurons [
6]. Additionally, recent work by Sommer et al. using CD3, a pan-T lymphocyte marker, revealed an increase in total CD3
+ T lymphocytes (including parenchymal and perivascular cells) in the substantia nigra of PD cases compared to controls [
68]. Our results corroborate these earlier findings and we also show that this increase is predominantly seen in PD dementia cases compared to controls. Furthermore, we observed a similar non-significant trend in the amygdala, particularly in the numbers of CD4
+ but not of CD8
+ T lymphocytes, and found that significant CD4
+ T lymphocyte infiltration (more than 10 cells) into the amygdala was more common in PDND and PDD (50% and 33% of cases) compared to controls (10% of cases). Infiltration of T lymphocytes in the brain parenchyma has also been observed in other synucleinopathies. In particular, CD4
+ but not CD8
+ or B lymphocytes were found to be increased in the frontal cortex and hippocampus of cases with dementia with Lewy bodies (DLB) compared to controls at postmortem [
35]. A second recent study in DLB cases showed increased T lymphocyte infiltration in both the grey and white matter of the middle temporal gyrus, in the absence of prominent microglial activation [
2]. Similarly, in the substantia nigra of cases with multiple system atrophy compared to controls both CD4
+ and CD8
+ T lymphocytes were found to be increased [
75]. The role of infiltrating T lymphocytes in PD is still unclear, however, ablation of CD4
+ T cells in an MPTP mouse model of PD was found to be neuroprotective [
6]. In another set of experiments using an AAV-α-synuclein rat model of PD it was observed that T cell-deficient (athymic nude) mice were protected from dopaminergic neuron loss in the substantia nigra [
70]. Taken together, this data suggests that these adaptive immune cells may have a cytotoxic effect in PD and related synucleinopathies. Furthermore, recent evidence from human studies suggests that α-synuclein epitopes are recognised by autoreactive CD4
+ T lymphocytes in PD [
71], which may explain our observed a significant correlation between CD4
+ T cells and α-synuclein pathology in the amygdala in our PD cases.
Both activated microglia and infiltrating lymphocytes may be exerting neurotoxic effects via the production of pro-inflammatory cytokines. In this study, we report for the first time an upregulation of the pro-inflammatory cytokine IL-1β in the frontal cortex of PD cases compared to controls. Gene expression of IL-1β was also increased in the PD substantia nigra, in line with previous evidence [
55]. It should, however, be noted that caution is needed when interpreting these results, given that the control sample size available for gene expression analysis was small. Furthermore, bulk tissue was used in these experiments with normalisation against housekeeping genes. Therefore, the reported gene expression findings have not been adjusted for potential differences in the ratio of neurons to glial cells which may occur due to increased neuron loss in certain regions in the PD cases compared to controls. The expression of pro-inflammatory cytokines in postmortem PD has not been extensively investigated in the past. In fact, previously available data come primarily from early work by Mogi and colleagues who quantified the protein levels of several cytokines in the substantia nigra using enzyme-linked immunoassays. They reported higher levels of both IL-1β and IL6 [
55], as well as elevated TNFα and IL2 in the substantia nigra of PD brains compared to controls [
56,
57].
One likely pathway leading to upregulation and secretion of pro-inflammatory cytokines is that mediated by Toll-like receptor activation. In vitro experiments have shown that microglia can be directly activated by misfolded α-synuclein through both TLR2 [
44] and TLR4 [
22,
66]. Increased protein levels of TLR2 and TLR4 have previously been reported in the substantia nigra [
15,
67], and the caudate/putamen of PD compared to control brains [
16], and in addition, these receptors are upregulated in peripheral blood mononuclear cells of PD patients compared to controls [
16,
74]. Our work has now shown elevation of TLR4 expression in multiple brain regions including the substantia nigra, amygdala and frontal cortex of PD cases compared to controls. Notably, in these same brain regions in PD, we also observed increased expression of IL-1β, a downstream product of the inflammasome pathway which is triggered by TLR4 activation. Interestingly, TLR2 and TLR4 have also been implicated in other proteinopathies, including Alzheimer’s [
69] and Huntington’s disease [
73] raising the possibility of a common pathogenic mechanism across several neurodegenerative diseases. The concomitant increase in the gene expression of TLR4 and IL-1β in the substantia nigra and the frontal cortex suggests an involvement of the NOD-like receptor protein 3 (NLRP3) inflammasome in these regions, with TLR4 activation resulting in increased expression of pro-IL-1β as well as NLRP3 activation; in turn, NLRP3 inflammasome activation could be responsible for the cleavage of pro-IL1β to the mature protein. This hypothesis is supported by recent data showing an upregulation of the protein levels of the NLRP3 adapter protein ASC (apoptosis-associated speck-like protein containing a caspase recruitment domain) as well as cleaved caspase-1 in postmortem nigral samples of PD cases compared to controls [
26]. The same study also showed that inhibition of NLRP3 activation in a mouse model of PD (intrastriatal injection of α-synuclein pre-formed fibrils) could effectively mitigate motor dysfunction as well as dopaminergic neuron loss. Nevertheless, future studies are necessary to determine whether NLPR3 inflammasome activation is also occurring in other extranigral regions, particularly in the amygdala and the frontal cortex of PD cases versus controls.
In this study, we observed significant correlations between cortical α-synuclein Lewy pathology and the rate of cognitive decline during life. This finding corroborates previous evidence showing a robust correlation between cortical Lewy body burden and cognitive decline in PD using multivariate linear regression analyses [
1,
64]. Other investigators have used the presence of dementia as the primary outcome in a logistic regression model, showing that neocortical Lewy body burden is strongly associated with dementia in PD [
30,
38]. In keeping with our findings that tau and amyloid-β pathology did not differ between demented and non-demented PD cases, we did not find a correlation between either tau or amyloid-β in any of the examined brain regions and MMSE decline per year.
Although our data implicate neuroinflammation, particularly in the amygdala in PDD, we did not find a significant correlation between inflammatory changes in this region and cognitive decline during life. A previous study similarly found no correlation between HLA-DR
+ microglia in the substantia nigra and clinical parameters in PD, whilst the use of a different marker CD68 (indicative of microglial phagocytic activity) revealed a strong association between CD68
+ microglia and disease duration [
12]. The method of characterizing activated microglia in postmortem brain may be critical to revealing clinicopathological correlations.
Indeed, a limitation of our study and a major challenge in postmortem brain studies overall is the definition of “activated microglia”. Here we have used enlarged and amoeboid morphology to quantify activated microglia selectively. However, this is a subjective method and microglial morphology is not restricted into either ramified or amoeboid shapes but represents a continuum including a whole range of morphological phenotypes [
7]. Another caveat in our analysis is that the immunostaining of microglia was performed on thin brain sections (10 µm). Previous studies assessing phenotypic differences to classify microglia have done so in sections 30–40 µm-thick [
24,
48], with Kongsui and colleagues finding that the diameter of many microglia ranges between 40 and 50 µm [
45] suggesting that even thicker sections would be needed for morphological studies. Furthermore, although microglia had been generally considered to be a functionally homogeneous population, comprehensive RNA sequencing studies have found evidence of different microglial subtypes with distinct function and have identified markers which can help distinguish between them. For instance, a recent single-cell RNA sequencing study in postmortem brain tissue from Alzheimer’s disease cases discovered a subtype of “disease-associated microglia” (DAM), with a unique transcriptional and functional profile, characterized by high phagocytic activity and upregulation of specific markers such as TREM2 [
43]. Therefore, future work in postmortem brain could make use of additional markers such as P2RY12 and TMEM119 (homeostatic microglia) [
4], and TREM2 (DAM) to better understand the role of microglia in PD. Future work could also utilise comprehensive genome-wide expression analyses which have become possible through the use of protocols to isolate microglia from human brain tissue [
58] or by singe-cell RNA sequencing using bulk tissue, as has recently been done in Alzheimer’s disease postmortem brain [
14].
Strengths of our study include our cohort of clinically well-characterised cases enabling the correlation between the pathological markers with the clinical course of the disease during life, as well as the use of controls who are representative of the typical aged population as discussed earlier. Additionally, in this study we employed a digital image analysis approach to quantitatively evaluate the severity of protein pathology in the postmortem brain. Traditional pathology is based on semiquantitative scoring upon visual inspection [
53,
54] and is a useful method of assessing the distribution and overall protein pathology burden but it is inevitably subject to inter-rater variability and may also lack sensitivity particularly in identifying subtle differences in pathology severity. Digital quantification is a reliable alternative, useful for high throughput analysis, and can provide a more accurate quantitative measure of pathology severity [
18,
23]. The methodology used in the present study was based on work described by Dunn et al. who showed a strong correlation between the automated analysis and the conventional scoring methodology [
18].
In summary, this study demonstrates that dementia in PD is associated with increased neuroinflammation in the substantia nigra and amygdala at postmortem, involving microglial activation and the infiltration of T lymphocytes. We also report an upregulation of the pro-inflammatory cytokine IL-1β and upstream TLR4 in both the substantia nigra and extra-nigrostriatal regions in PD. We have confirmed that limbic and neocortical α-synuclein is the most robust predictor for dementia in PD and identified a correlation between α-synuclein and neuroinflammation in the amygdala. Taken together, this data suggests that a combination of α-synuclein pathology and inflammatory changes in the brain are critically involved in dementia in PD.