In this study, 189 patients with PD, 60 patients with ET, and 190 healthy controls were recruited from the Department of Neurology, Ruijin Hospital, Shanghai Jiao Tong University School of Medicine from March 2014 to December 2015. The clinical diagnosis of PD was established by a senior movement disorder specialist using the clinical criteria for PD (the United Kingdom PD Society Brain Bank clinical diagnostic criteria; Calne et al., 1992; Hughes et al., 1992). PD symptoms were evaluated using the Unified Parkinson’s Disease Rating Scale (UPDRS; van Hilten et al., 1994) and the modified Hoehn and Yahr scale (Hoehn and Yahr, 1967). In this study, no family history of parkinsonism was present in all PD patients. None of them carries DJ-1 L166P mutation, which is associated with autosomal recessive early-onset PD (Bonifati et al., 2003), and confers reduced protein stability of DJ-1 (Moore et al., 2003). The control samples, including healthy persons and ET patients, were matched to PD patients in gender, age, ethnicity, genetically unrelated, and had no family history of parkinsonism. All the subjects were of Chinese Han ancestry. The study was approved by the Ethics Committee of Ruijin Hospital affiliated with Shanghai Jiaotong University School of Medicine and the written informed consent was obtained from all participants. All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. The demographics and clinical characteristics of participants for biomarker study were summarized in Table 1.

Blood samples were collected in tubes containing EDTA. Plasma was separated within 4 h by centrifuging the blood at 3000× g for 10 min. Plasma was immediately separated into a 1.5 ml RNase-free centrifuge tube and stored at −80°C until RNA isolation. Three PD samples and five healthy control samples were selected for miRNA array analysis, and the remaining samples were saved for validation.

Plasma samples from PD patients and healthy controls were submitted to Kangchen Corporation (Kangchen, Shanghai, China) for miRNA microarray analysis. Total RNA was harvested from plasma using TRI reagent BD (MRCgene, TB-126) according to the manufacturer’s instructions. The RNA samples were labeled with the miRCURY™ Hy3™/Hy5™ Power Labeling Kit (Exiqon, Vedbaek, Denmark) and hybridized to the miRCURY™ LNA Array (v.18.0; Exiqon). miRNA expression data were normalized and chosen for differentially expressed miRNAs screening.

Microarray results were confirmed by qRT-PCR. For miRNA analysis, total RNA was extracted from plasma with TRIzol LS reagent (Invitrogen) and cDNAs of specific miRNAs were synthesized with the specific reverse transcript primers. Delta delta CT method was used to calculate miRNA expression. Plasma miRNA expression results were normalized to the expression of endogenous miRNA hsa-miR-191-5p.

RT primer for hsa-miR-191-5p:

For DJ-1 mRNA analysis, total RNA was extracted from cells with TRIzol reagent (Invitrogen) and DJ-1 mRNA expression levels were normalized to GAPDH.

QPCR primers for human DJ-1:

The following antibodies were used: rabbit polyclonal anti-PARK7/DJ-1 antibody (Abcam, ab18257, Cambridge, UK), mouse monoclonal anti-β-actin antibody (Sigma-Aldrich, Clone AC-15, St. Louis, MO, USA), horseradish peroxidase (HRP)-conjugated goat anti-mouse IgG, goat anti-rabbit IgG (Jackson ImmunoResearch Laboratories, PA, USA). 5-(and-6)-chloromethyl-2′,7′-dichlorodihydrofluorescein diacetate, acetylester (CM-H2DCFDA) reagents were purchased from Invitrogen (CA, USA). Dihydroethidium (DHE) reagents were purchased from Beyotime (Jiangsu, China). All chemicals were purchased from Sigma-Aldrich (St. Louis, MO, USA) unless otherwise stated.

miRNA mimics and inhibitors were purchased from Ribobio Company (Guangzhou, China). Universal negative controls for mimics and inhibitors were based on the sequences of C. elegans, which has been confirmed to have minimal homeology to all known miRNAs of miRBase 18.0. All of the sequences of miRNAs are from the Sanger miRNA database¹.

SH-SY5Y cells were seeded in a 96-well-plate at the number of 5000/well. Twenty-four hours after transfecting miRNA mimics/inhibitors, cells were treated with neurotoxic substance MPP⁺ (0.5 mM) or rotenone (0.15 μM) for another 24 h. Ten microliters of CCK-8 (Obio Technology, Shanghai) was added to the culture medium, and the viability of the cells was measured 4 h later at 450 nm with a Microplate reader (Synergy Mx, Bio-Tek, Winooski, VT, USA) according to the manufacturer’s instructions.

Data from three or more independent experiments are presented as the mean ± SD. Statistical analysis were done using SPSS 22.0 and GraphPad Prism 5.0 softwares. Student’s t-test, one-way analysis of variance (ANOVA), or Kruskal-Wallis test were used for statistical analysis. Correlations between miRNA levels and clinical characteristics were tested with Spearman’s rank correlation analysis. Differences with p < 0.05 were defined as the threshold for significance.

Information about the complete methods is provided in Supplementary Material. We used the same methodologies, such as cell culture and transfection, western blotting, dual-luciferase reporter assay and reactive oxygen species (ROS) measurement, as the ones employed in our previous study (Xiong et al., 2014).

Global plasma miRNA expression profiles were measured between the early PD group (Hoehn and Yahr stage 1–2.5) and healthy control group by miRNA microarray. The global miRNA expression patterns between the two groups were significantly different (Figure 1). Analysis of miRNA expression profiles showed 50 miRNAs were significantly dysregulated vs. controls, including 31 up-regulated miRNAs and 19 down-regulated miRNAs (fold change >1.5 or < −1.5, P-value <0.05; Figure 1). Next, seven miRNAs (hsa-miR-34c-3p, hsa-miR-148b-5p, hsa-let-7i-3p, hsa-miR-4639-5p, hsa-miR-34a-3p, hsa-miR-181a-5p and hsa-miR-30a-5p) were picked for validation with plasma samples from 20 sporadic PD patients and 20 healthy controls. The results showed similar expression tendency with that of microarray, except hsa-miR-34a-3p (Figure 2), indicating that microarray data was reliable in our study.

To better understand the roles of observed dysregulated miRNAs and the underlying mechanism in PD pathogenesis, we utilized several prediction tools including Targetscan (Grimson et al., 2007), miRanda (Betel et al., 2008) and miRWalk (Dweep et al., 2011) to predict biological targets of above-mentioned miRNAs. Among a number of target gene candidates, we were particularly interested in PD-related genes, which have clear links with the pathogenesis of the disease. Notably, hsa-miR-4639-5p was predicted to relate to DJ-1 (PARK7), which plays a crucial role in cellular oxidative stress response (Shendelman et al., 2004). In PD patients, decreased DJ-1 protein levels have been observed in several brain areas (Kumaran et al., 2009; Nural et al., 2009; Miñones-Moyano et al., 2011). DJ-1 deficiency could lead to severe mitochondria fragmentation (Irrcher et al., 2010) and ROS burst, contributing to increased sensitivity to oxidative stress and apoptosis (Demasi and Davies, 2003). Based on these studies and our prediction data, we wondered if hsa-miR-4639-5p indeed participates in the regulation of DJ-1 expression. Thus, hsa-miR-4639-5p was chosen for further study.

In our preliminary data, hsa-miR-4639-5p was observed to be abnormally up-regulated in PD patients. In order to mimic such pathological change, hsa-miR-4639-5p was overexpressed by the transient transfection of hsa-miR-4639-5p mimics, and the level of DJ-1 protein was analyzed by western blotting. HEK-293T cells were used in experiments, which express significant amounts of endogenous DJ-1. Overexpression of hsa-miR-4639-5p resulted in a dose-dependent decrease in DJ-1 protein level (Figure 3A). Luciferase assay was used to further determine whether DJ-1 is a putative target for hsa-miR-4639-5p. Luciferase reporter construct was generated with the 3′UTR of human DJ-1 mRNA cloned after Renilla luciferase coding sequence. The construct was co-transfected with hsa-miR-4639-5p mimics or negative control. Accordingly, hsa-miR-4639-5p was observed to repress the reporter activity markedly (Figure 3C, black bar). These results suggest that hsa-miR-4639-5p might be the human DJ-1 expression regulator.

Seed sequence at nucleotides 2–8 of the 5′ end of miRNA is fundamental for target recognition and binding (Bartel, 2009). We found that the seed sequences of hsa-miR-4639-5p are perfectly complementary to nucleotides 217–224 of human DJ-1 3′UTR. RNAhybrid (Krüger and Rehmsmeier, 2006), which is designed for miRNA target prediction, was used to test our result. The minimum free energy required for interaction between hsa-miR-4639-5p and the corresponding DJ-1 3′UTR binding site is −18.3 kcal/mol according to RNAhybrid calculation (Figure 3B). To confirm direct binding of hsa-miR-4639-5p to DJ-1 3′UTR, predicted binding sites were mutagenized as shown in Figure 3C. hsa-miR-4639-5p was not able to repress luciferase activity when mutagenized 3′UTR was applied, indicating that the predicted sequence is indeed the genuine binding site for hsa-miR-4639-5p.

To further prove the effects of abnormal up-regulated hsa-miR-4639-5p on DJ-1 expression, miRNA inhibitors (synthetic single-stranded anti-sense oligonucleotides designed to inhibit miRNA activity) were co-transfected with hsa-miR-4639-5p mimics into HEK-293T cells. hsa-miR-4639-5p inhibitors resulted in restored DJ-1 expression in a dose-dependent manner (Figure 3D). Luciferase activity of human DJ-1 was also significantly restored in the presence of hsa-miR-4639-5p inhibitors (Figure 3E). Meanwhile, alteration of DJ-1 protein levels by transfecting hsa-miR-4639-5p mimics or inhibitors was confirmed in human neuroblastoma SH-SY5Y cells (Supplementary Figure S1).

Furthermore, there was no significant change of the levels of DJ-1 mRNA in the cells transfected with hsa-miR-4639-5p mimics (Figure 3F), indicating that hsa-miR-4639-5p repressed DJ-1 protein level through affecting translation of DJ-1 mRNA.

Our previous study (Wang et al., 2011) suggested that DJ-1 is able to protect cells against oxidative insult. DJ-1 knock-down cells treated with MPP⁺ produced more ROS and were more vulnerable to oxidative stress than those of normal controls. Here, we investigated the effect of abnormal up-regulated hsa-miR-4639-5p on cellular anti-oxidative function when cells are exposed to stressful attacks. SH-SY5Y cells were transiently transfected by 50 nM miRNA mimics or 50 nM miRNA mimics + 50 nM miRNA inhibitors, followed by MPP⁺ or rotenone treatment (both of which are toxic substrates that induce oxidative stress). Twenty-four hours later, ROS levels and cell viability were measured (Figure 4A). The results showed that hsa-miR-4639-5p-overexpressed SH-SY5Y cells elicited more ROS and O2− than the control cells or miRNA inhibitors co-transfected cells (Figures 4B,C). Moreover, cell viability decreased in hsa-miR-4639-5p-overexpressed cells compared to control cells or miRNA inhibitors co-transfected cells (Figure 4D).

To further investigate whether abnormal up-regulated hsa-miR-4639-5p impaired anti-oxidative function through its direct regulation on DJ-1 expression, DJ-1 protein was re-introduced into hsa-miR-4639-5p-overexpressed cells. We constructed two plasmids that included or did not include 3′UTR of human DJ-1 gene, namely, DJ-1CDS-3′UTR and DJ-1CDS. miRNA mimics + DJ-1CDS-3′UTR plasmid or miRNA mimics + DJ-1CDS plasmid were co-transfected into cells (Figure 4E). hsa-miR-4639-5p was able to inhibit exogenous DJ-1CDS expression when it’s 3′UTR was existed. However, it could not inhibit exogenous DJ-1 expression when only DJ-1CDS was transfected. The results showed that ROS production and cell death decreased in DJ-1CDS co-transfected cells compared to DJ-1CDS-3′UTR co-transfected cells (Figures 4F–H).

These results suggested that cells with abnormal up-regulated hsa-miR-4639-5p were more susceptible to MPP⁺ or rotenone-induced cell death as a consequence of DJ-1 depletion.

To further confirm the abnormal up-regulation of hsa-miR-4639-5p level in PD patients, we performed qPCR analysis of plasma hsa-miR-4639-5p level in a relatively large cohort of patients with PD (n = 169), healthy controls (n = 170) and patients with ET (n = 60). The demographics information of all participants is summarized in Table 1. The relative hsa-miR-4639-5p level in PD patients was significantly higher (P < 0.001) than that in healthy controls and ET patients (Figure 5A). Receiver operating characteristic (ROC) curve reflects a separation between the PD and normal subjects, with area under the curve (AUC) values up to 0.939 (95% CI = 0.914–0.965, sensitivity = 93.49%, specificity = 85.29%, Figure 5B).

Further, we wondered whether plasma hsa-miR-4639-5p level could serve as an early diagnosis marker for PD. Thus, PD patients were divided into sub-groups according to Hoehn and Yahr stage and disease duration, and stratified analysis was performed. Plasma hsa-miR-4639-5p level was found to be significantly up-regulated in early stage of PD and maintained relatively high levels during PD progression (Figures 5C,D). ROC curve showed that hsa-miR-4639-5p has adequate sensitivity and specificity to discriminate between early PD patients and healthy controls (PD patients with H-Y 1–2.5 vs. NC: sensitivity = 95.28%, specificity = 85.88%; PD patients with disease duration ≤2 years vs. NC: sensitivity = 100.00%, specificity = 85.88%; Figures 5E,F). Moreover, plasma hsa-miR-4639-5p level was uncorrelated with gender, age of disease onset, levodopa daily equivalent dose (LDED), and UPDRS score of PD patients (Supplementary Figure S2). Therefore, the elevated plasma level of hsa-miR-4639-5p in PD patients is stable.