Cerebrospinal Fluid Biomarkers in Parkinson’s Disease: A Critical Overview of the Literature and Meta-Analyses

Parkinson’s disease (PD) is a common neurodegenerative disorder; however, well-established biochemical markers have not yet been identified. This review article covers several candidate cerebrospinal fluid (CSF) biomarkers for PD based on the recent literature and meta-analysis data. The decrease of α-synuclein in PD is supported by meta-analyses with modest reproducibility, and a decrease of amyloid β42 is seen as a prognostic marker for cognitive decline. Tau, phosphorylated tau (p-tau), and neurofilament light chains have been used to discriminate PD from other neurodegenerative disorders. This article also describes more hopeful biochemical markers, such as neurotransmitters, oxidative stress markers, and other candidate biomarkers.


Introduction
Parkinson's disease (PD) is one of the most common neurodegenerative disorders, and is characterized by resting tremors, bradykinesia, rigidity, and postural instability, as well as cognitive symptoms [1]. Pathologically, this disorder is defined by degeneration of the substantia nigra and the presence of Lewy bodies, which contain abundant amounts of α-synuclein. Researchers have also found that gene mutations of α-synuclein cause familial forms of PD. Therefore, α-synuclein is considered to play a major role in the pathogenesis of PD.
The current clinical diagnostic criteria of PD are based on existence of parkinsonism and the exclusion of other disorders, and all the tests for PD, such as magnetic resonance imaging or nuclear medicine imaging are positioned as supportive tools [2]. Therefore, good biomarkers for PD with high sensitivity and specificity are desired.
The cerebrospinal fluid (CSF) biomarkers for PD have been investigated widely to elucidate the pathophysiology and to support differentiating PD from normal subjects or other neurological disorders, such as multiple system atrophy (MSA) [3] or progressive supranuclear palsy (PSP) [4]. CSF biomarkers are also sought for the evaluation of disease activity/progression, severity, staging, comorbidity (i.e., cognitive dysfunction), and prognosis.
Confirmed CSF biomarkers have not yet been established; however, some candidate molecules (for example, α-synuclein) have been inspected as potential ones. In this article, we aimed to provide a comprehensive overview for the biomarkers for PD and referred to recent meta-analyses for several biomarkers.

Materials and Methods
This study is based on a literature search in the database PubMed with the keywords (Parkinson AND cerebrospinal fluid AND biomarker AND (review OR meta-analysis)) on 24 May 2020. Non-English publications were excluded from this review because of the difficulty of assessing Of the 255 articles identified in the database, 63 were included in this study ( Figure 1). The search also found 15 meta-analysis articles referring to α-synuclein, neurofilament light chains, cytokines (interleukin (IL)-1β, IL-6, and transforming growth factor (TGF)-β1), amyloid Aβ42, tau, and phosphorylated tau.

Materials and Methods
This study is based on a literature search in the database PubMed with the keywords (Parkinson AND cerebrospinal fluid AND biomarker AND (review OR meta-analysis)) on 24 May 2020. Non-English publications were excluded from this review because of the difficulty of assessing the contents as available open resources. Only the most relevant or earliest studies were included in this review and duplicated contents were omitted as well as studies with a small sample size (n < 20).
We also focused on the following markers in particular: (1) α-synuclein and its related molecules; (2) dopamine, neurotransmitters, and the metabolites; (3) oxidative stress markers; (4) amyloid beta as a predictor for cognition; and (5) tau, phosphorylated tau, and neurofilament light chains as discriminators for atypical Parkinsonism.

Results
The results are summarized in Table 1 (the biomarkers for distinguishing PD from controls), Table 2 (the biomarkers for distinguishing PD from other neurodegenerative disorders), and Table 3 (the biomarkers for representing disease severity or cognitive abilities). Then, we made detailed comments for each marker category.

Results
The results are summarized in Table 1 (the biomarkers for distinguishing PD from controls), Table 2 (the biomarkers for distinguishing PD from other neurodegenerative disorders), and Table 3 (the biomarkers for representing disease severity or cognitive abilities). Then, we made detailed comments for each marker category.

Neurotransmitters and Neuromodulators: Focusing on Dopamine and the Metabolites
The depletion of dopaminergic neurons in the substantia nigra is an essential pathology in PD, and noradrenergic neurons in locus ceruleus were also observed. Dopamine and noradrenaline decreases are also potential candidates of PD, as well as the metabolites, including dihydroxyphenyl acetic acid (DOPAC) and homovanillic acid (HVA). However, the levels of these compounds actually have different sources and meanings [7]. CSF homovanillic acid is rather distantly related to neuronal dopamine stores and reflects several intervening processes. As dopaminergic neurons do not contain catechol-O-methyltransferase, CSF homovanillic acid depends on the uptake and intracellular O-methylation in extra-dopaminergic cells. Thus, in PD, the striatal content of homovanillic acid is not as severely decreased as that of dopamine. CSF dopamine also may not provide an accurate reflection of central dopamine deficiency. Dopamine in extracellular fluid is derived mainly from exocytotic release in response to pathway traffic and the escape of neuronal reuptake by the cell membrane dopamine transporter.
As dopaminergic neurons are lost, pathway traffic to the remaining terminals likely increases compensatorily, thereby augmenting dopamine delivery from those terminals to the extracellular fluid; thus, CSF dopamine may underestimate the extent of the loss of neuronal dopamine stores. CSF dopamine concentrations are infinitesimal-sometimes below the detection limit of the assay method [7]. Meta-analysis articles were not found in the PubMed database in this category. CSF catecholamine would be influenced by levodopa intake, and, therefore, the measurements should be performed for a levodopa-naïve or washed-out state.

Growth Factors
Only sparse data was available for this area. Brain-derived neurotrophic factor (BDNF) was significantly different (decreased) between the control and both neurodegenerative groups (PD and Alzheimer's disease) but not between neurodegenerative groups [25]. CSF progranulin was investigated in PD, amyotrophic lateral sclerosis and controls, but no difference was found among the groups [30].

α-Synuclein and Its Related Molecules
α-synuclein is considered one of the most important targets in this field, and many studies have investigated CSF α-synuclein. The reported control/normal CSF α-synuclein values varied widely; however, meta-analyses showed that the CSF levels of α-synuclein decreased in patients with PD [31][32][33][34]. One possible reason for the conflicting results is the contamination of CSF by α-synuclein from the blood, as α-synuclein is abundant in whole blood, plasma, and serum, in which its levels are up to 10 2 -10 3 times higher than those found in the CSF [64]. Synuclein oligomer and phosphorylated α-synuclein are also candidate molecules supported by a meta-analysis [34]. The meta-analyses showed that the sensitivity and specificity of α-synuclein were 0.72-0.88% and 0.40-0.65, respectively [32,34]. The sensitivity and specificity of α-synuclein oligomer were 0.71 and 0.64, respectively [34] (Table 4). Specificity 85% (95% CI 79-89%) ↑ = increase, ↓ = decrease, → = no change. AD, Alzheimer's disease; APD, Atypical parkinsonian disorders; CI, confidence interval; IL, interleukin; MSA, multiple system atrophy; PD, Parkinson's disease; PDCI, Parkinson's disease with cognitive impairment; PDNC, Parkinson's disease with normal cognition; PSP, progressive supranuclear palsy; SMD, standardized mean difference; TGF, transforming growth factor; WSD, weighted mean difference. * Please note this SMD value represents comparison of AD to PD as described in the reference [33].
A recent study demonstrated that CSF α-synuclein decreased early in the disease, preceding motor PD. CSF α-synuclein does not correlate with the progression and therefore does not reflect ongoing dopaminergic neurodegeneration. Decreased CSF α-synuclein may be an indirect index of changes in the balance between α-synuclein secretion, solubility, or aggregation in the brain, reflecting its overall turnover [65]. This corresponds with another finding that the Hoehn-Yahr stage was not correlated with the CSF α-synuclein level, and that the striatal binding ratio on dopamine transporter imaging with 123 I-ioflupane decreased in the PD group, but this was not correlated with the CSF α-synuclein level [66]. Another area of interest regarding CSF α-synuclein is its relationship with cognitive function; however, the research remains inconclusive or less influential [66]. Technical caution and pitfalls for the measurement of α-synuclein should be considered for appropriate use. Perianalytical considerations for biomarker studies in PD have been summarized [67]. α-synuclein is abundant in erythrocytes and, therefore contained in the serum, plasma, and whole blood, and, thus, blood contamination should be avoided cautiously. CSF hemoglobin is often referred to as a marker for blood contamination. The use of traumatic needles, discarding the first five drops of CSF, and collection with polypropylene or siliconized tubes are recommended for sampling for CSF. There does not appear to be any diurnal fluctuation.
Why α-synuclein decreases in the CSF of PD remains unknown. Low production, aggregation in the brain, and/or increased clearance of α-synuclein would be possible.

Amyloid Beta, Tau, and Phosphorylated Tau as Predictors for Cognition
Cognitive decline in PD is well known, and several studies reported a relationship between amyloid-beta and cognition. Another recent topic of research interest is the relationship between PD, amyloid β42 (Aβ42), tau, and p-tau, particularly from the viewpoint of cognitive function. One meta-analysis showed that the CSF Aβ42 level in PD with the cognitive impairment (CI) cohort was lower than that in PD with the normal cognition (NC) cohort. Reduced Aβ42 as well as elevated t-tau and p-tau were observed in the PD with dementia (PDD) cohort compared with the PDNC cohort. Therefore, the meta-analysis concluded that amyloid pathology and tauopathy may participate in the development of PDD, which is similar to Alzheimer's disease [35].
It is unknown whether α-synuclein and amyloid beta can interact; however, a previous study reported no significant correlation between α-synuclein and amyloid beta [66]. There is no robust evidence regarding the contribution of α-synuclein to cognitive function. These matters should be investigated in further research.

Neurofilament Light Chains for Discriminating PD from Other Neurological Disorders
Neurofilament light chains (NFL) also have the potential to differentiate PD from MSA or PSP, and this is supported by meta-analyses [57][58][59][60]. NFL elevation was observed in MSA, PSP, and corticobasal degeneration. In terms of the comparison between PD and atypical parkinsonian disorders, the sensitivity and specificity of NFL were 0.82 and 0.85, respectively [59] (Table 4).
Recent studies focused on the hypothesis that epigenetic mechanisms contribute to PD development and progression. Epigenetics refers to the regulatory mechanisms of gene expression that are not mediated by the DNA sequence itself, but by chemical or allosteric DNA modifications or by the action of regulatory noncoding RNAs. MicroRNAs (miRNAs) are small noncoding RNAs that serve as posttranscriptional regulators of gene expression. They bind to messenger RNA (mRNA) and promote mRNA degradation and/or decrease the translation. One meta-analysis identified four studies for miRNA of CSF; however, no significant signals were found in the CSF, although the study Brain Sci. 2020, 10, 466 9 of 14 identified 13 significantly differentially expressed miRNAs in the brain (n = 3) and blood (n = 10) [51]. The most compelling finding was miRNA hsa-miR-132-3p binding to the mRNA of the α-synuclein gene (SNCA).

Discussion
Cumulative data, including meta-analyses, support that a decrease of CSF α-synuclein is observed in PD and that a decrease of amyloid β42 is a predictor of cognitive decline in PD. Tau, phosphorylated tau, and neurofilament light chains were able to discriminate PD from other neurodegenerative disorders. The elevation of cytokines (IL-1β, IL-6, and TGF-β) in PD suggested pro-inflammatory mechanisms in the pathogenesis of PD. CSF DOPAC appeared to provide a sensitive means to identify PD. This review could not identify reliable CSF oxidative stress markers in the literature thus far.
Further studies are necessary to reveal α-synuclein abnormalities, which may include the transcription, translation, epigenetic regulation, and kinetics (production, distribution, post-translational modification, oligomerization, aggregation, clearance, etc.). The standardization of measurement techniques is also required to use the practical application of these markers.
This review has some limitations. First, the abovementioned meta-analyses showed statistical significance; however, heterogeneity was present in each meta-analysis. Overlapping of the data between PD and the counterpart is present in any marker; therefore, they should be applied carefully. Second, these CSF biomarker samplings were performed mainly while the patients were alive, and, therefore, the diagnosis at the time of CSF sampling was established with clinical criteria but not as a histopathological diagnosis.
It is inevitable to have the limitation of pre-mortem differential diagnosis of AD and dementia with Lewy bodies (DLB). Certain studies adopted post-mortem CSF sampling with neuropathologically diagnosed PD subjects [18]; however, post-mortem changes may happen in the CSF or even in the brain, and the post-mortem data should be compared with longitudinal information. Certain patients showed a mixed pathology of AD and Lewy pathology [68]. The employment of other surrogate markers, such as amyloid positron emission tomography (PET), tau PET, or dopamine transporter imaging, would be necessary for the verification of the clinical diagnosis, as well as other modalities, such as myocardium metaiodobenzylguanidine (MIBG) and dopamine transporter (DAT) scintigraphy. Longitudinal data sampling is also needed but limited [65].

Conclusions
In conclusion, the meta-analyses for cerebrospinal fluid biomarkers of PD showed: (1) a decrease of α-synuclein was a marker of PD, (2) a decrease of amyloid β42 was a marker of cognitive decline in PD, (3) the elevation of tau, phosphorylated tau, and neurofilament light chains differentiated between PD and the related disorders, and (4) the elevation of cytokines (IL-1β, IL-6, and TGF-β) was observed in PD. Many other candidate biomarkers were also scrutinized. These findings will aid in the accurate diagnosis of PD and other neurodegenerative disorders and facilitate understanding of the pathogeneses of these conditions. Further studies will be needed to obtain more precise measurements of the biochemical biomarker levels. Any single biomarker cannot lead to a 'snap-shot' diagnosis. Combinations of several biomarkers would help with more accurate diagnoses and evaluation of the pathophysiology.
CSF biomarkers would aid in understanding the pathophysiological mechanisms, improve the diagnostic accuracy, and facilitate the future development of novel therapies, including disease-modifying drugs.
Finally, we refer to the limitation of this article. This work is a narrative review but not a systematic review. Therefore, we admit that this article could contain some selection bias although each report was fairly scrutinized as much as possible. The aim of this work is to provide the latest comprehensive review regarding cerebrospinal fluid biomarkers of PD, to grade the reliance of each marker according to meta-analyses, and to focus on each marker or marker category with respect to future research directions. We agree that the systematic review method is a more reliable scientific approach. However, referring to all reports offsets the conciseness of this article, and even dopamine and the metabolites, which are generally regarded as central issues in the pathophysiology of PD, have no meta-analysis yet. Hence, we did not adopt the systematic review approach in this article.