Parkinson’s disease (PD) is the second most common neurodegenerative disease characterized by progressive loss of dopaminergic neurons in substantia nigra. While hereditary PD is mainly caused by gene mutations such as in SNCA, Parkin, and PINK1, the majority of this disease is sporadic and the pathogenesis involves both genetic vulnerabilities and environmental exposures (Bekris et al., 2010). Thus far, diagnosis of PD largely relies on the presence of motor symptoms, clinically characterized by bradykinesia, resting tremor, rigidity, and at late stage gait instability. However, the actual onset of this disease precedes the motor manifestations by a number of years (Kalia and Lang, 2015). As a consequence, possible clinic or preventive interventions lag far behind the disease onset. Substantial efforts have therefore been made to develop potential markers for early diagnosis of PD, and the strategies include prodromal clinical signs, imaging, skin or colonic biopsies, genetic sequencing, and biochemical testing in cerebral spinal fluid, blood, saliva, and urine. For example, certain non-motor symptoms such as olfactory impairment and rapid eye movement sleep behavior disorder, which are frequently present in PD before motor manifestations, have been under investigation as potential markers for PD diagnosis (Postuma et al., 2012). Unfortunately, thus far none of them has been proven effective in clinical trials despite early promises (Kalia and Lang, 2015).

MicroRNAs (miRNAs) act as endogenous regulators of gene expression by binding to complementary sequences of target messenger RNA (Chen and Rajewsky, 2007). Being small non-protein coding RNAs and also relatively stable, disturbance of miRNA in tissues is possibly reflected by changes in serum/plasma, which renders miRNA a potentially convenient biomarker in monitoring diseases (Reid et al., 2011). miRNAs have also been suggested to play an important role in PD and profiled in brains of PD patients (Kim et al., 2007; Minones-Moyano et al., 2011). Among the miRNAs known to be associated with PD, in particular those with postmortem evidence and/or regulating α-synuclein expression, are: miR-34b and miR-34c, down-regulated in the brains of PD patients (Minones-Moyano et al., 2011; Villar-Menendez et al., 2014) and repressing α-synuclein expression (Kabaria et al., 2015); miR-7 and miR-153, suppressing α-synuclein expression (Junn et al., 2009; Doxakis, 2010); miR-433, regulating fibroblast growth factor 20 (FGF20) levels and subsequent α-synuclein expression (Wang et al., 2008); and miR-133b, down-regulated in midbrain of PD patients and interacting with Pitx3 (Kim et al., 2007). In this study, we focused on these six miRNAs to better understand PD-associated changes and explore their potentiality as PD biomarkers.

Deficiency in efficient biomarkers hinders timely intervention of PD and thus demands continuous endeavors. Aberrant miRNA expression has been suggested in a strength of linkage to sporadic PD (Harraz et al., 2011). Herein, we demonstrate that plasma levels of miR-433 and miR-133b significantly differ between PD patients and controls, following screening a set of cell-free PD-associated miRNAs. These two miRNAs exhibit strong correlations in terms of their expression levels in plasma, and share several common signaling pathways.

miR-433 binds to the promoter region of FGF20. A variant, rs12720208, at the miRNA binding site of FGF20 is associated with PD, as its risk allele prevents miR-433 from attaching and leads to increased FGF20 translation and subsequent α-synuclein expression (Wang et al., 2008). The reduced expression of miR-433 as suggested in plasma may contribute to increased FGF20 and α-synuclein expression, assuming that the plasma level reflects the changes in the brain. In accordance with our results, a recent profiling of extracellular miRNAs by small RNA sequencing also suggests a significant decrease of miR-433 in cerebrospinal fluid (CSF) of postmortem PD patients. Their profiling in serum shows five differentially expressed miRNAs (Table 3), but no statistical difference was found in miR-433 and miR-133b (Burgos et al., 2014). One possible reason for the discrepancy is that the profiling samples are from postmortem patients, which usually are at very late stage of the disease. Meanwhile, specific analysis by qPCR may be more sensitive than general sequencing profiling.

The miR-133b is enriched in dopaminergic neurons and is deficient in midbrain tissues of PD patients (Kim et al., 2007). The miR-133b suppresses midbrain dopaminergic neuron maturation and function by targeting Pitx3 as disclosed in primary rat midbrain cultures (Kim et al., 2007). The deficiency of miR-133b in PD is further confirmed using laser microdissected postmortem dopaminergic neurons in a case–control study (Sonntag, 2010). In accordance, miR-133b level is significantly decreased in plasma/serum of PD patients as shown in the current study and a previous report (Zhao et al., 2014). Interestingly, different from the above in vitro model (Kim et al., 2007), miR-133b null mice exhibit normal numbers of midbrain dopaminergic neurons during development and aging along with unchanged Pitx3 expression and dopamine release in the striatum (Heyer et al., 2012). This result indicates that miR-133b per se is not sufficient to induce the pathogenesis of PD, or its reduction in PD may merely be a consequence of dopaminergic neuronal loss during disease progression.

In contrast to miR-433, miR-133b, miR-34b, and miR-153, which are stably measurable in plasma, miR-34c and miR-7 are present at very low level in plasma, indicating that these two miRNAs may not be suitable to be biomarkers. On the other hand, our limited sample volume prevents us from detecting these low abundant miRNAs through the use of higher amount of RNA for the original cDNA synthesis. Results from microarray and sequencing analyses show that miR-34c and/or miR-7 are expressed in peripheral blood mononuclear cells (PBMCs) and in serum, although their expression not altered in PD (Martins et al., 2011; Burgos et al., 2014).

Pairwise correlations between miR-433, miR-133b, miR-34b, and miR-153 showed a trend of reduction in the PD group, indicating a differential involvement of these miRNAs in PD pathogenesis and/or pathology. Notably, the strongest correlation is observed between miR-433 and miR-133b, the two miRNAs with significant changes in PD. It appears that aging has no impact on levels of these miRNAs in plasma. However, further studies are needed from populations of wider age range since the age of participants in this study resides in a narrow range. It is surprising that changes in miR-433 and miR-133b expression are not associated with PD severity, although this seems not a sole case (Zhao et al., 2014; Li et al., 2017). Further studies are needed with larger sample size for each stage.

Potential target genes of miR-433 and miR-133b were involved in various pathways, sharing seven in common. However, no overlap was found in these genes, neither in the biological processes, suggesting that they function likely in a complementary manner. Modeling using data of miR-433 and miR-133b does not generate satisfactory performance for PD prediction. It may attribute to our relative small sample size that limits the cross-validation to build an optimal machine-learning model. On the other hand, it may not be surprising since a single form of biomarker has been considered almost impossible to predict PD and a combination of multiple testing including imaging, biochemical, and genetic is proposed to be required for future success (Kalia and Lang, 2015).

Proper normalization is critical for detection of authentic biological differences between samples. Although there is no consensus on the miRNA loading control (Laurent et al., 2015), many studies have used nucleolar RNA U6 as an internal control to normalize circulating miRNAs (Ai et al., 2010; Wang et al., 2013; Zhao et al., 2014), while some studies simply follow a consistent reaction system using no normalization controls as noted earlier (Chen et al., 2008; Ren et al., 2013). Besides, endogenous miR-16-5p, considered to be similar among individuals in most cases, was occasionally used as an internal control (Song et al., 2012). However, serum or plasma prepared from hemolyzed blood samples may contain miR-16-5p that can be increased by 20- to 30-fold (Pritchard et al., 2012). Exogenous synthesized miRNAs such as Caenorhabditis elegans miR-39/54/238 are also used for normalization (Mitchell et al., 2008; Brase et al., 2011; Tan et al., 2014), which, however, lack the capacity to normalize biological and pathological variations (Laurent et al., 2015). In general, higher amount of qPCR template can reduce the cycle number and obtain lower Ct value for potentially more stable values, and larger sample size helps the power of analysis (Dong et al., 2016). Unfortunately, there is realistic challenges to acquire more volume and more samples such as in the study by Muller et al. (2014). In the current study, some Ct values are relatively high in measurements of miR-433 and miR-153. As indicated in Section “Materials and Methods,” triplicate repeats are used during each qPCR to help the quality control of reliability. The normalization methods using U6 as an internal control and using no control lead to similar results, suggesting that our reactions are stably performed following a consistent system.

As summarized in Table 3, miRNA biomarkers of PD are investigated in blood, PBMCs, leukocytes, plasma, serum, and CSF. There seems no report in urine and saliva, although cases are found in other diseases such as cancer (Park et al., 2009; Yamada et al., 2011). Interestingly, the identified differentially expressed miRNAs in PD are varied among studies (Table 3). As mentioned earlier, we selected the six miRNAs based on postmortem evidence in brain and/or regulating α-synuclein expression. Not considering the studies which measure miRNAs individually selected by their respective rationales, results from screenings by arrays or next generation sequencing are surprisingly rarely overlapped (Martins et al., 2011; Khoo et al., 2012; Cardo et al., 2013; Soreq et al., 2013; Botta-Orfila et al., 2014; Burgos et al., 2014; Vallelunga et al., 2014; Gui et al., 2015; Ding et al., 2016; Dong et al., 2016). This discrepancy may be attributed to differences in ages, disease stages and subtypes of included patients, threshold of significance in data processing, sample sizes, and type of sources if using different fluids/cells. Of note from the overall results, a decreased level of miR-19b/miR-19b-3p in PD is suggested by four studies, respectively, in PBMCs (Martins et al., 2011), serum (Botta-Orfila et al., 2014; Cao et al., 2017), and CSF (Gui et al., 2015), except one study which shows increased miR-19b-3p expression in CSF (Burgos et al., 2014). The expression of miR-195 is suggested increased in serum of PD by two studies (Ding et al., 2016; Cao et al., 2017). The expression of miR-30c is decreased in PD, respectively, in PBMCs (Martins et al., 2011) and serum (Vallelunga et al., 2014). Decreased levels of miR-133b and miR-433 in plasma of PD as shown in the current study are in line with two previous reports, respectively in serum and CSF (Burgos et al., 2014; Zhao et al., 2014). Also, consistent with our results, miR-153 and miR-34b are not reported to be changed in fluids of PD. miR-34c and miR-7, which are not sufficiently detected in our condition, are not suggested to be changed either in PD (Table 3). Further investigations are warranted to clarify how levels of miRNAs change in the fluids of PD in association with PD subtypes and disease progression. Also, the differences may be worth testing in urine and saliva in order to further understand their potentiality to serve as PD biomarkers.

While it is a long haul to eventually identify clinic proven biomarkers, we herein screen and demonstrate that levels of miR-433 and miR-133b, two miRNAs associated with PD pathogenesis, are significantly reduced in plasma of PD patients and not correlated with age, suggesting their specificity and potentiality to serve as biomarkers. Future studies are needed in combination with other testing to upgrade their performance in predicting PD.

J-HZ, XZ, and H-MW designed and supervised the project. XZ, B-LH, and Z-YH collected samples and clinical data. RY and L-LZ performed research. RY, PL, and H-MW analyzed data. J-HZ and RY wrote the article.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.