Elevated plasma miR-133b and miR-221-3p as biomarkers for early Parkinson’s disease

Blood circulating microRNAs (miRNAs) are proposed to be promising biomarkers for many neurodegenerative disorders, including Parkinson’s disease (PD). However, there is a lack of identified differentially expressed miRNAs in PD from different studies. The aim of this study was to evaluate miRNAs expression in PD. We measured plasma circulating miRNA expression in three independent sets with a total of 151 PD patients, 21 multiple system atrophy (MSA) patients and 138 healthy controls using high-throughput RT-PCR. We identified that elevated miR-133b and miR-221-3p discriminated early-stage PD from controls with 94.4% sensitivity and 91.1% specificity. Elevated miR-133b and miR-221-3p distinguished PD from controls with 84.8% sensitivity and 88.9% specificity. In addition, miR-4454 distinguished PD from MSA with 57.1% sensitivity and 82.6% specificity. Hence, elevated miR-133b and miR-221-3p potentially represent good biomarkers for early PD, and a combination of miR-133b, miR-221-3p and miR-4454 has the potential to serve as a non-invasive biomarker for PD diagnosis.


Results
Screening of candidate miRNAs in PD. Based on the S-Poly(T) Plus miRNA assay, qRT-PCR was utilized to detect human plasma miRNA profiles (486 miRNAs) between 78 PD patients and 78 controls with three pools (Fig. 1). The demographic characteristics of the cases are shown in Table 1. In this study, exogenous cel-miR-54-5p was added during the RNA extraction process as a normalization control. Then, highly and consistently expressed miRNAs were selected as candidate endogenous reference genes. The stability of the candidate reference genes was further evaluated by two different algorithms (geNorm and NormFinder). Hsa-miR-25-3p was finally found to be quantifiable and stable in all samples, demonstrating no significant differential expression between groups (p > 0.05) and less variation between inter-and intra-group. Hence, hsa-miR-25-3p was used as an endogenous reference gene in this study. All the following assessments were double-checked with cel-miR-54-5p and Hsa-miR-25-3p as exogenous and endogenous reference miRNAs, respectively, and the results from these two reference miRNAs were independently consistent. Hence, all data presented in this study were normalized to cel-miR-54-5p.
Analysis of miRNAs profile showed 32 miRNAs significantly dysregulated in PD compared with controls, including two down-regulated (fold changes ≤ 0.65, p ≤ 0.05) and 30 up-regulated (fold changes ≥ 1.5, p ≤ 0.05). The results were summarized in Table 2. Among these 32 miRNAs, miR133b 13,14 , miR-320 15,16 , miR-221-3p 14 , miR-627-5p 17 and miR-205 18 had been proposed to involve in the pathogenesis of PD from literature review or bioinformatic analysis. In addition, they were also shown in relative low Ct value (≤ 33) ,large fold changes (> 2 or < 0.65, Table 2), and p value < 0.05 in screening test, indicating as good candidate miRNAs. Considering four miR-320 families (miR-320a/b/c/d) were recruited from the screening set, the miR-320a, which was proposed to have the strongest relationship with PD 15,16 and presented with the largest change fold and lowest Ct value in this family (change fold = 3.03, Ct value = 23, Table 2), was selected as a symbol for miR-320 family. Furthermore, miR-432-5p and miR-4454 were also selected as candidates as they had the largest change folds or lowest Ct value among these 32 miRNAs with good p value (Table 2). Hence, seven miRNAs (miR-432-5p, miR-133b, miR-320a, miR-4454, miR-221-3p, miR-627-5p and miR-205) were selected as candidate miRNAs for further validation.

Duplication of miRNAs biomarkers in PD.
Twenty-seven PD cases (including 13 L-dopa-naïve PD and 14 L-dopa-treated PD) and 15 age-and sex-matched controls were recruited in this set. There were no statistically significant differences in age, sex, disease duration, modified Hoehn and Yahn (H-Y) stage or UPDRS scores were noted between the L-dopa-naïve and L-dopa-treated PD groups. The demographic characteristics of the cases are shown in Table 1.
L-dopa naïve PD versus L-dopa treated PD. As previously reported, L-dopa and other drug treatments may alter miRNA expression in PD 19 . Hence, we first aimed to determine whether PD drug treatments would affect the expression of seven candidate miRNAs. Statistically significant differences in the seven candidate miRNA levels were not noted between the L-dopa-naïve and L-dopa-treated PD groups (t test, p > 0.05, Supplementary Fig. 1), indicating that L-dopa treatments probably have minimal effect on the expression of these candidate miRNAs.
PD versus controls. Furthermore, the levels of seven candidate miRNAs were assessed between 27 PD cases and 15 controls. The levels of miR-133b, miR-221-3p and miR-320a were significantly higher in PD cases compared to controls (for miR-133b, 2.04 times increase, t test, p = 0.004; for miR-221-3p, 2.47 times increase, t test, p < 0.001; for miR-320a, 1.68 times increase, t test, p = 0.014; Fig. 2). There was no significant difference www.nature.com/scientificreports/ in the expression of miR-205, miR-432-5p, miR-4454 and miR-627-5p between the PD group and controls (t test, p > 0.05). No age-or sex-dependent difference in the expression of the seven candidate miRNAs was found between the control or PD groups. Analyses of covariance between groups with sex and age as covariates showed sex and age did not exert significant effects on the considered miRNAs. The diagnostic accuracy of miRNAs differentially expressed between PD and controls was further assessed by ROC curve analysis. The area under the curve (AUC) value for miR-221-3p was 0.80, with 74.1% sensitivity and 86.7% specificity (Fig. 3a). The AUC value for miR-133b and miR-320a was 0.74 with 77.8% and 66.7% sensitivity and 66.7% and 86.7% specificity, respectively (Fig. 3b,c). Combining these three miRNAs increased the AUC value to 0.82. They sensitivity was 81.5%, and the specificity was 86.7% (Fig. 3d). The diagnostic power of miRNAs to discriminate between PD and healthy controls is summarized in Table 3.
Biomarkers for early PD diagnosis. Based on modified H-Y staging, PD cases were divided into early stages (modified H-Y stage 1-2.5) and advanced stages (modified H-Y stage 3-5, Table 1). To find biomarkers for early PD diagnosis, 18 PD cases in early clinical stages were selected for comparison with 15 controls (Table 1). The levels of miR-133b and miR-221-3p were significantly upregulated in early-stage PD compared to controls (for miR-133b, 1.95-fold increase, post-hoc with Bonferroni, adjust p = 0.05; for miR-221-3p, 2.09-fold increase, post-hoc with Bonferroni, adjust p = 0.007; Fig. 4a,b). No significant differences on the levels of other five candidate mRNA between early-stage PD and controls were found. For early PD diagnosis, the AUC values for miR-133b and miR-221-3p were 0.68 and 0.75, respectively. The sensitivity and specificity of miR-133b were both 66.7%, and those of miR-221-3p were 72.2% and 86.7%, respectively (Fig. 4c,d). The combination of miR-133b and miR-221-3p did not change the AUC value (0.75) and the sensitivity was 77.8%%, and the specificity was www.nature.com/scientificreports/ 80.0% (Fig. 4e). The diagnostic power of miRNAs to discriminate between early-stage PD and healthy controls is summarized in Table 3.

Validation of predictive miRNA biomarkers in PD. PD versus controls versus MSA.
To validate the diagnostic value of seven miRNA candidates, a new independent cohort with 46 PD, 21 MSA and 45 healthy controls was assessed (Table 1) and analyzed. The levels of miRNA-133b, miRNA-221-3p and mi R-4454 were significantly different among control, PD and MSA groups. MiR-133b expression was significantly upregulated in both PD and MSA compared to controls (3.67-fold increase in PD, post-hoc of Kruskal-Wallis analysis, adjust p < 0.001; 2.70-fold increase in MSA, post hoc of Kruskal-Wallis analysis, adjust p < 0.001; Fig. 5a&c). No significant difference in miRNA-133b expression was noted between PD and MSA (post-hoc of Kruskal-Wallis analysis, adjust p = 1.00, Fig. 5a). Meanwhile, miRNA-221-3p levels were significantly upregulated in PD but not in MSA compared to the control (for PD, 1.79-fold increase, post hoc of Kruskal-Wallis analysis, adjust p = 0.032; Fig. 5b). Although miR-4454 expression was not significantly different between the PD and control groups, an unexpectedly moderate upregulation of  Fig. 5c). There was no significant difference in miR-205, miR-320a, miR-627-5p and miR-432-5p expression among PD, MSA and controls (Kruskal-Wallis analysis, p > 0.05, supplementary Fig. 2). There was no significant difference in miR-320a expression between PD and controls (p = 0.058), which was different from the result in duplication set. Higher miRNA-221-3p expression was found in female compared to that in male in However, these sex-based differences were not observed in control or MSA group. No sex-dependent difference in the expression of other candidate miRNAs were found in three groups. Analyses of covariance between groups with sex and age as covariates showed sex and age did not exert significant effects on the candidate miRNAs.
To investigate the diagnostic value of the candidate miRNAs for PD, logistic regression analysis was performed. Univariate logistic regression found that increased plasma miR-133b (p < 0.001), and miR-221-3p (p = 0.012) differentiated PD patients from healthy controls ( Table 4). All factors with a p value < 0.05 were subsequently included in the multivariate logistic regression (stepwise) analysis, which demonstrated that miR-133b expression (p < 0.001) and miR-221-3p (p = 0.041) might also be independent factors for diagnosis of PD (Table 5). ROC curve analysis revealed good diagnostic values for miR-133b to differentiate PD from controls (AUC values 0.84, 82.6% sensitivity and 88.9% specificity; Fig. 6a), whereas the diagnostic values for miR-221-3p were lower (AUC values 0.64, 55.2% sensitivity and 80.0% specificity; Fig. 6b). Combining miR-133b with      Fig. 6c). Interestingly, ROC curve analysis of miR-133b also showed good diagnostic value to distinguish MSA from controls (AUC values 0.88, 85.7% sensitivity and 84.4% specificity; Fig. 6d). The AUC value for miR-4454 to distinguish PD from MSA was 0.72with 57.1% sensitivity and 82.6% specificity (Fig. 6e). The diagnostic power of miRNAs to discriminate among PD, MSA and healthy controls is summarized in Table 5.
Biomarkers for early PD diagnosis. To identify biomarkers for early PD diagnosis, 18 PD cases in early clinical stages were selected for comparison with 45 controls in the validation set (Table 1). We also observed dramatical increases in the levels of miR-133b and miR-221-3p in early-stage PD compared to controls (for miR-133b, 4.21fold increase, post-hoc of Kruskal-Wallis analysis, adjust p < 0.001; for miR-221-3p, 1.74-fold increase, post-hoc of Kruskal-Wallis analysis, adjust p = 0.035; Fig. 7a,b). The expression of seven candidate miRNAs in different PD stages and controls is shown in Fig. 7 and Supplementary Fig. 3. For early PD diagnosis, the AUC values for miR-133b and miR-221-3p were 0.90 and 0.67, respectively. The sensitivity and specificity of miR-133b were 94.4% and 91.1%, respectively, and those of miR-221-3p were 55.6% and 80%, respectively (Fig. 7c,d). The combination of miR-133b and miR-221-3p did not change the diagnostic value (AUC value 0.90, 94.4% sensitivity and 91.1% specificity, Fig. 7e). The diagnostic power of miRNAs to discriminate between early-stage PD and healthy controls is summarized in Table 5.

Discussion
Previous studies have identified several miRNAs that are potentially involved in the pathogenesis of PD 2,5,14 ; however, the application of circulating miRNAs as biomarkers for PD remains debated 2,5 . Blood and its derivatives are the most extensively studied sources of biomarkers. Previous studies have indicated that plasma is more suitable than serum in studying circulating miRNA, as RNA released during the coagulation process may change the true repertoire of circulating miRNA 10 . However, plasma miRNAs are present at lower concentrations than serum miRNAs, thereby necessitating higher requirements for detection methods. In this study, we measured plasma circulating miRNA expression in three independent sets with a total of 151 PD, 21 MSA and 138 healthy controls using high-throughput RT-PCR. We identified elevated miR-133b and miR-221-3p levels that differentiated early-stage PD from controls with good diagnostic value. Moreover, we confirmed that a combination of miR-133b, miR-221-3p and miR-4454 distinguished PD from controls and MSA.
L-Dopa and related PD medications have been shown to dramatically affect the expression of certain miR-NAs in PD 19 . In this study, we did not find any difference in the expression of seven candidate miRNAs between the drug-treated group and drug-naïve group, minimizing the contamination of miRNA expression by drug treatment in the following studies. Sex-based differences of miRNAs have been reported in human brains as Table 4. Diagnostic power of miRNAs to discriminate among PD, MSA and healthy controls (measured by ROC curve analyses). PD, Parkinson's disease; MSA: multiple system atrophy; AUC, area under the curve; CI, confidence interval.  20,21 . In this study, we compared seven candidate miRNAs expression between male and female in both duplication set and validation set, and failed to find stable sex-based difference. In addition, analyses of covariance between groups with sex and age as covariates showed sex and age did not exert significant effects on the considered miRNAs, minimizing gender-induced bias. The next comparison between PD and controls showed remarkable upregulation of plasma miR-133b, miR-221-3p and miR-320a in PD, and elevated miR-133b and miR-221-3p levels discriminated early-stage PD from controls with good diagnostic value. MiR-133b is one of the first reported miRNAs playing roles in the maintenance of midbrain dopaminergic neurons in both cell and animal models 13 . MiR-133b is specifically enriched in the midbrain and deficient in PD compared to the control 13 . Reduced serum circulating miR-133b has also been observed in PD patients compared to controls 22,23 . Interestingly, high-throughput small RNA sequencing in a Drosophila PD model presents consistent upregulation of miR-133 in early-stage PD files 24 . The precise pathophysiological functions of miR-133b in PD are not completely understood. Overexpression of miR-133b in primary embryonic rat midbrain culture reduced dopaminergic neuron numbers and dopamine release, whereas its suppression had the opposite effect, suggesting that it could be a negative regulator of dopaminergic neurons 13 . Hence, the simplest explanation for miR-133b deficiency in the PD brain and its toxic mouse model is the loss of dopamine neurons where miR-133b is normally enriched 25 . In this study, we identified significant upregulation of plasma circulating miR-133b expression in both the early-stage and advanced-stage PD groups independent of L-dopa treatment. These findings are consistent with the limited effect of PD drug treatment on miR-133b expression 19 . Surprisingly, miR-133b was also upregulated in MSA compared to controls, suggesting that miR-133b may also be related to α-synuclein pathologies in addition to dopaminergic neurons. A recent report inferred that miR-133b indirectly influences α-synuclein by targeting the Ras homolog gene family member A (RhoA) 26 . MiR-221-3p is another upregulated miRNA identified in PD in this study. To our knowledge, at least five previous studies have assessed miR-221 expression in PD (Table 3) [27][28][29][30][31] . Interestingly, miR-221 expression in brain tissue is remarkably increased in PD 27,28 , and elevated cingulate miR-221 correlates with SNCA and PARK2 transcript levels 28 . However, the alterations in serum miR-221 levels in PD are variable in different studies [29][30][31] . These discrepant results from brain tissue and serum have not yet been completely www.nature.com/scientificreports/ explained. MiR-221 is highly expressed in glioma cells and implicated in various biological processes, including apoptosis, cell cycle and differentiation 32 . Cytobiological studies have shown that miR-221 may be involved in the pathogenesis of PD by regulating autophagy 33 , oxidative pathways 34 and iron uptake 35 . To date, there has been no assessment of circulating plasma miR-221 expression in PD. In the present work, we performed triplicate measurements in three independent sets and identified that elevated plasma miR-221-3p predicts both early and advanced PD in controls. In addition, increased plasma miR-221-3p expression was only observed in PD but not in MSA compared to controls, indicating a potential differentiation value of miR-221-3p between PD and MSA. Previous studies have shown that miR-320a inhibits cell proliferation and regulates tumor occurrence, progression and metastasis. MiR-320a is decreased in the blood samples of colorectal cancer 36 and increased in the sera of patients with Alzheimer's disease(AD) 37 . Interestingly, a previous study also showed a reduction in miR-320a in blood leukocytes of PD patients compared to controls 15 . In this study, we observed the upregulation of plasma miR-320a in the advanced stage of PD compared to healthy controls in duplication set. However, the increases of miR-320a in PD group could not be replicated in the validation set. The underlying mechanism by which miR-320a contributes to neurodegeneration remains to be investigated. Overexpression of the miR-320a mimic has been shown to precipitate α-synuclein accumulation by suppressing heat shock protein 70 16 . Combined with findings of altered miR-320a expression in AD 37 , PD and MSA, we propose that miR-320a may have a close relationship to synucleinopathies and be involved later in disease progression than miR-133b. But the instable alternations of miR-320a in duplication set and validation set in this study doubt its real role in PD and more studies with larger sample size are needed to identify this result. MiR-4454 is a relatively new micro-RNA with limited investigation. Most current studies of miR-4454 focus on cancers 38 , and no qualification of miR-4454 expression in PD has been reported. In the present study, miR-4454 levels were unchanged in global PD compared to controls in both the duplication set and validation set. However, plasma miR-4454 levels in PD patients were lower than those in MSA patients, suggesting that miR-4454 could be a potential supplementary biomarker for differentiating PD from other neurodegenerative disorders.  Table 6. We noticed there are some discrepancies. This difference may be explained by several reasons. First, different sample sources and measurement methods from different studies may lead to conflicting results. Second, since miRNAs are tissue-and disease-specific, there is a lack of universally accepted reference miRNAs to which target miRNAs can be normalized. U6 RNA and 5S rRNA, which are normally used as housekeeping miRNAs for tissue miRNA normalization, are degraded in serum samples 29 . Finally, the storage time of samples should also be considered 39 .
There are a few limitations in this study. First, all samples in the current study were from mono-center in China, and further international multiple center studies are needed to avoid selection and racial bias. Second, the alternations of miR-320a in duplication set and validation set were inconsistent, showed the challenges of miR-320a as a biomarker for PD diagnosis. Third, only limited MSA samples were assessed as disease controls in this study. More neurodegenerative diseases are needed to verify the diagnostic value of the detected miRNAs in the future.
In summary, this work revealed that upregulated plasma miR-133b and miR-221-3p potentially represent good biomarkers for early PD, and a combination of miR-133b, miR-221-3p and miR-4454 has the potential to serve as a noninvasive biomarker for PD diagnosis.

Materials and methods
Participants and experiments design. The study was approved by the Ethics Committee of Xiangya Hospital (2013Q09). All individuals were informed of the use of blood samples and signed a written informed consent before the start of the study. A total of 310 participants with 151 PD, 21 MSA and 138 healthy controls were recruited in this investigation for three independent test sets. All PD patients met the diagnostic criteria of the Movement Disorder Society in 2015 40 , and all MSA patients were diagnosed according to Gilman's criteria 41 . In addition, all PD and MSA cases were confirmed by two independent senior neurologists. Participants with other diseases, such as cancer, diabetes, cardiovascular disorders, infection and any other neurological or psychiatric disorders, were excluded.
In the initial screening set, 78 PD cases and 78 healthy controls were randomly divided into 6 groups (26 samples of each group) for plasma miRNA profiling. In the duplication test set, 27 PD cases (including 13 L-dopanaïve PD and 14 L-dopa-treated PD) and 15 healthy controls were included for candidate miRNA assessments. In the validation set, 46 PD, 21 MSA and 45 healthy controls were recruited.
Sample collection and circulating RNA extraction. Five milliliters of peripheral whole blood were collected from each participant in EDTA anticoagulation tubes and kept at room temperature for 1 h. Then, blood samples were centrifuged at 3000 rpm and 4 °C for 15 min. The supernatants were plasma, aliquoted and stored at − 80 °C. Circulating RNAs were isolated from plasma by the miRsol method as described in our previous study 11 . Caenorhabditis elegans cel-miR-54-5p (0.1 pM) was added as a normalization control. MiR-451a and U6 RNA were used as quality controls to exclude cell and hemolysis contamination.
Polyadenylation, reverse transcription. The polyadenylation and reverse transcription procedure were conducted exactly according to the S-Poly(T) Plus protocol 11,42 . To profile miRNAs effectively, every 7 out of 486 miRNAs as well as spiked-in cel-miR-54 were grouped for the one-step reaction of polyadenylation and reverse transcription simultaneously. MiRNAs with identical forward primers or with more than five base pairings between the forward primer and RT primers were avoided in the same group. All sequences, primers and probes are listed in Supplemental Table 1. www.nature.com/scientificreports/ Quantitative real-time PCR. qRT-PCR was performed in 96-well plates by using an ABI Step-One Plus thermal cycler as previously described 11 . Each PCR was performed in duplicate. The miRNA expression level was normalized to spiked-in cel-miR-54-5p. MiRNAs with cycle threshold (Ct) values less than 35 in the panel were included in the data analysis.
Statistical. All analyses were performed with SPSS 22.0 (SPSS Inc., Chicago, IL, USA). Data were presented as the mean ± SEM. A paired t-test was used to identify candidate miRNAs in the screening step. Independent t-tests were used to assess differences between groups in the duplication set, and nonparametric analysis of the Mann-Whitney U test was used to assess differences between groups in the validation set. Univariate and multivariate logistic regression analyses were used to assess the diagnostic value of candidate miRNAs. A receiver operating characteristic (ROC) curve was used to calculate the relationship between sensitivity and specificity for the disease group versus healthy or disease controls and hence evaluate the diagnostic performance of the analyses, either individually or in combination. The "optimum" cutoff value from the ROC curve was determined by the sum of sensitivity, and specificity was maximal. p value < 0.05 were considered to be significant.
Informed consent. Informed consent was obtained from all subjects involved in the study.
Ethical approval. All procedures performed in studies involving human participants were in accordance with the ethical standards of the ethical committee of the Xiangya Hospital, Central South University, China, and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

Data availability
The data that support the findings of this study are available from the corresponding author upon reasonable request.