Association of the serum microRNA-29 family with cognitive impairment in Parkinson's disease

We aimed to examine whether miRNA-29s (miR-29s) in serum are associated with cognitive impairment in Parkinson’s disease (PD). Thirty-nine PD patients with normal cognition (PD-NC), 37 PD patients with mild cognitive impairment (PD-MCI), 22 PD patients with dementia (PDD) and 40 healthy controls were recruited. Detailed clinical evaluations and a schedule of neuropsychological tests were administered to all patients. MiR-29s expression in serum samples was assessed using reverse-transcription quantitative real-time PCR. We found that the levels of all three miR-29s in the PDD group were significantly lower than those in the PD-NC group (p < 0.05). In addition, the miR-29b level was downregulated in the PD-MCI group with respect to that in the PD-NC group (p < 0.05). After adjusting for years of education and the UPDRS-III subscore using a multivariate model, miR-29s showed significant associations with PDD. MiR-29b levels were shown to be associated with different subsets of PD cognition and could accurately discriminate PDD from non-PDD (area under the curve (AUC) = 0.859; 95% CI, 0.7817-0.9372). Further analysis of the cognitive domains found that the miR-29s levels were all associated with memory performance in PD patients. In summary, miR-29s are associated with cognitive impairment in PD.


INTRODUCTION
Cognitive impairment (CI) is the most prevalent and debilitating nonmotor symptom in Parkinson's disease (PD) [1], ranging from mild cognitive impairment (PD-MCI) to dementia (PDD). However, the current diagnosis of PDD, based on both the DSM-IV criteria and the proposed clinical diagnostic criteria, is time-consuming and somewhat subjective [2]. Reliable and easily applicable biomarkers for PDD or even earlier stages of impairment, such as PD-MCI, are urgently needed.
MicroRNAs (miRNAs) are single-stranded sets of [21][22] nucleotides that act as posttranscriptional modulators by inhibiting or promoting the degradation of their mRNA targets [3,4]. The potential of miRNAs as peripheral biomarkers of dementia, such as Alzheimer's disease (AD) [5], vascular dementia [6] and frontotemporal dementia (FTD) [7], is being actively pursued and shows promise. It has been reported that miRNAs contribute to cognitive impairment by playing critical roles in neurodevelopment, synaptic plasticity, memory and the regulation of neurodegenerative disease-associated pathological proteins [8,9]. However, to our knowledge, no study has previously investigated the roles of miRNAs in cognitive impairment in PD.
We previously reported that the members of the miR-29 family (miR-29a, miR-29b and miR-29c), a group of brain-specific miRNAs, were downregulated in the serum of PD patients and showed a decreasing trend related to more severe Parkinsonism [10]. Considering that cognitive performance worsens with increasing disease severity in PD [11] and the critical role of miR-29s in neuronal survival [12], aging [13], and synaptic plasticity [14], we wondered whether miR-29s were associated with cognitive deficits in PD. To explore this, we compared serum miR-29s expression in PD patients with different cognitive states and evaluated their potential diagnostic accuracy for PDD and PD-MCI patients.

Clinical characteristics and neuropsychological tests of participants
We enrolled 98 PD patients and 40 healthy controls (HCs) in total. The patients were classified into 3 groups: patients with PDD (n = 22), patients with PD-MCI (n = 37), and patients with PD-NC (n = 39). The four groups were gender-and age-matched (p > 0.05; Table 1). No group differences were found with respect to ESS scores, SSST-12 scores or RBDSQ scores among the PD patients (all p > 0.05). The PDD group, followed by the PD-MCI group, had the longest disease duration, the most sever motor impairment (H&Y, UPDRS-III), and the highest GDS scores and hallucination incidence, while having the shortest years of education. The LEDs in the PDD and PD-MCI groups were both higher than those in the PD-NC group.
Unsurprisingly, the scores on the MMSE and specific cognitive assessments were dramatically different among the PDD, PD-MCI and PD-NC patients. The patients' detailed cognitive profiles are shown in Table 2.

MiR-29s and cognitive impairment in PD patients
MiR-29a/b/c levels showed no differences between the PD-NC group and HC group (p>0.05). MiR-29a/b/c expression was significantly lower in the PDD group than those in the PD-NC group (Figure 1) (p<0.05). The expression levels of miR-29b and miR-29c were both significantly lower in the PDD group than those in the PD-MCI group (p<0.05). Additionally, miR-29b expression in the PD-MCI patients was significantly downregulated with respect to that in the PD-NC patients (p<0.01).
We built a univariate model (Model 1) to explore the "crude" effect of miR-29s on incident PDD or PD-CI. As shown in Table 3, the univariate logistic regression suggested significant associations between the expression of all members of the miR-29s family with PDD and PD-CI (PDD + PD-MCI). These associations remained significant after adjusting for years of education and UPDRS-III subscore in a multivariate model (Model 2) with the exception that miR-29c levels were not associated with PD-CI.

MiR-29s and clinical characteristics of PD patients
MiR-29s levels showed no correlation with the clinical parameters of interest, including age, gender, years of education, disease duration, LED, H&Y stage, UPDRS-III subscore, ESS score, GDS score, SSST-12 score, RBDSQ score or the incidence of hallucination (Table 5).

DISCUSSION
The present study compared serum miR-29a/b/c levels in PDD, PD-MCI, PD-NC patients and healthy controls. Low expression of miR-29s in PD patients was significantly associated with poorer cognitive function.   In particular, miR-29b differentiated PDD from non-PDD with high accuracy and PD-CI from PD-NC with moderate accuracy.
Biomarkers that allow early identification of cognitive dysfunction have the potential to assist in predicting the onset of dementia in PD and could possibly lead to earlier interventions. Previous research has made efforts to identify biomarkers of PD-CI, including attempts to show relationships with the levels of Aβ in CSF, uric acid in plasma/serum, measurements of cerebral cholinergic innervation and metabolism using PET, and hippocampal size measured by MRI [15]. Either the invasive nature or high costs of these techniques preclude their routine use in large populations. This study is an attempt to widen the list of potential biomarkers to include blood-based miRNAs. Although many miRNAs have been identified to be associated with PD [16], to the best of our knowledge, this is the first study to directly explore whether miRNAs in blood could be peripheral biomarkers for cognitive impairment in PD.
miRNAs have been reported as peripheral biomarkers of other neurocognitive disorders, such as Alzheimer's disease (AD) [5], vascular dementia [6] and frontotemporal dementia (FTD) [7]. Upregulation of miR-29c was shown to promote learning and memory behaviors in AD mice [17]. In this study, we illustrated downregulation patterns of serum miR-29s in PD with cognitive impairment. Specifically, the expression levels of miR-29s, particularly miR-29b, decreased in the PDD and PD-MCI groups compared with those in the PD-NC group. Furthermore, miR-29b, while having a high AUC in discriminating PDD from non-PDD and a moderate AUC in discriminating PD-CI from PD-NC, showed no correlation with other clinical parameters. After adjusting for other confounders, miR-29b continued to be associated with global cognitive parameters. The specific association with cognitive impairment makes miR-29b a potential candidate biomarker for distinguishing PDD patients.
In specific domains, we noted that the levels of the miR-29s family were all associated with the z-score of memory function. As PD-NC progresses to PDD, studies have shown that memory deficits become apparent, implicating a supervening dysfunction of temporal lobe storage mechanisms [18]. This association between miR-29s and memory function may result from the high expression of miR-29s in the central nervous system, particularly in neurons of the hippocampus [14]. Additionally, miR-29a and miR-29b showed an association with the language zscore, and miR-29b expression contributed to executive deficits, but the underlying mechanisms need to be further explored.
Our previous study showed a marked reduction in serum miR-29s levels in PD patients compared with those in the unaffected controls. In this study, we further divided PD patients into three groups according to cognitive status. However, serum miR-29s levels in the PD-NC patients and healthy controls showed no significant differences, but serum miR-29s levels in the PD-MCI and PDD patients were significantly lower than those in healthy controls. This might indicate that miR-29s are more related to cognition but not pure Parkinsonism. These results may be due to the different neurobiological bases underlying cognitive and motor deficits in PD. Additionally, our previous study found that serum miR-29s did not differ between AD patients and healthy participants [10]. It seems that serum miR-29s are associated with PD-CI but not AD.
Currently, the mechanisms underlying cognitive impairment in PD remain unclear [19]. miR-29s may provide insights into these pathological mechanisms.   The target genes for mature miR-29a/b/c heavily overlap according to computer prediction. The synaptic regulator PARK7 (DJ-1), mitogen-activated proteins MAPK6, MAPK7, and MAPK10, short-term memory to long-term memory conversion regulator CREB, and neurogrowth and neurotrophic factors IGF1 and IGF2 are candidate targets of miR-29s. The pathogenetic role of miR-29s in PDD or PD-MCI warrants further study.
Our study has some limitations. First, this study has a small sample size. Thus, a larger group of patients is needed to confirm these results. Second, this study utilizes a cross-sectional design, which could not be used to analyze the longitudinal impacts of miR-29s on cognitive function in PD. Longitudinal studies are required to explore whether miR-29s could identify PD patients at risk for further cognitive decline. Additionally, our cross-sectional study could not make causal inferences, and the role of miR-29s in the pathogenesis of cognitive impairment in PD is worthy of further study.
In conclusion, we identified the association of the serum miR-29 family with cognitive impairment in PD. Our findings reveal that the serum microRNA-29 family, especially miR-29b, may be potential biomarkers for PDD and PD-CI.    [20]. Patients with any history of stroke, epilepsy, encephalitis, traumatic brain injury, malignancies, cardiac events, or severe psychiatric illness were excluded from the study.
Forty age-and gender-matched control subjects were voluntarily recruited. None of the control subjects had a history of neurologic/psychiatric disorders.

Clinical evaluation
All participants went through clinical assessment after at least 12 h off anti-parkinsonian medications. Motor symptom evaluation included the Unified Parkinson's Disease Rating Scale Part III (UPDRS-III) and the modified Hoehn and Yahr staging (H&Y), while nonmotor symptoms were measured by the Epworth Sleepiness Scale (ESS), the Geriatric Depression Rating Scale (GDS), the Sniffin' Sticks Screening 12 Test (SSST-12) [21], and the Rapid Eye Movement Sleep Behaviour Disorder Screening Questionnaire (RBDSQ). Total daily levodopa equivalent doses (LEDs) were calculated to represent the doses of PD medications.
All participants were in the ON condition during cognitive assessment to minimize the confounding impact of motor symptoms. To obtain normative data for the Chinese adult population, we recruited 100 healthy controls matched for age, education, and gender (Supplementary Table 1). The raw score of the individual neuropsychological tests was transformed into a z-score by subtracting the mean test score of the control sample from the individual raw scores and then dividing the difference by the standard deviation of the score of the control sample according to the following formula: control control Z score (test score Mean )/SD . = − The mean of two z-scores for each domain was used as the compound score.
PDD was diagnosed by the current clinical diagnostic criteria of the Movement Disorder Society (MDS) Task Force 2007 [30,31]. PD-MCI was defined based on the MDS Task Force 2012 (Level 2) criteria [32], and impairment (>1.5 SD below the normative mean) on at least 2 neuropsychological tests within the same cognitive domain or across different domains was required. PD-CI includes both PDD and PD-MCI. Non-PDD includes both PD-MCI and PD-NC.

Reverse-transcription quantitative real-time PCR (RT-qPCR) for miRNA
Serum samples were collected from each participant on the same day as the clinical and neuropsychological assessments. Serum miRNAs were extracted using a miRNeasy Serum/Plasma Kit (Qiagen, Germany), during which proportional miRNeasy Serum Spike-In Control was added as the reference RNA. Reverse transcription was performed with a miRcute miRNA First-Strand cDNA Synthesis Kit (Tiangen, China), followed by quantitative real-time PCR of 2 µl of the product. The PCR primer sequences (purchased from Life Technologies) were as follows: miR-29a (5'-TAG CACCATCTGAAATCGG-3'); miR-29b (5'-TAGCA CCATTTGAAATCAGT-3'); and miR-29c (5'-TAGCA CCATTTGAAATCGG-3'). Relative expression levels were calculated using the 2-ΔΔCt method and normalized to the level of the reference RNA. The whole RT-qPCR process was assessed by two experienced researchers blinded to the clinical and neuropsychological data.

Data analysis
Continuous variables are shown as the means ± SDs, while categorical data are presented as numbers and frequencies (%). Data were analyzed using IBM SPSS Statistics (version 21). P < 0.05 was defined as statistically significant.
For quantitative data, differences among the three PD groups were compared by one-way analysis of variance (ANOVA) with an LSD post hoc test and the Kruskal-Wallis test for normally distributed and nonnormally distributed data, respectively. Categorical data were analyzed with the chi-squared test or Fisher's exact test as appropriate. To test the diagnostic accuracy of miR-29s, we used receiver operating characteristic curve (ROC) analyses.
We calculated correlations between miR-29s expression levels and clinical parameters (age, years of education, H&Y, UPDRS-III, etc.) using Pearson's (r) correlation or Spearman's (rho) correlation as appropriate. A univariate model and a multivariate model were used to explore the effect of miR-29s expression levels on incident PDD or PD-CI with or without considering other confounders (including years of education and UPDRS-III subscore). Controlling for the same covariates, independent associations of miR-29b expression levels with cognitive domain zscores were evaluated by multivariate linear regression analysis.

Ethics statement
This investigation was conducted in accordance with the ethical standards of the Declaration of Helsinki and according to national and international guidelines, and has been approved by the Human Studies Institutional Review Board, Huashan Hospital, Fudan University. All participants provided written informed consent.

AUTHOR CONTRIBUTIONS
Dr. JW and Dr. FH had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Y-LT, FH and JW conceived and designed the study. L-LH, X-CB, YF and YS performed experiments. L-LH and X-NL analyzed and interpreted data. L-LH and Y-LT drafted the manuscript. FH and JW revised the manuscript. All authors read and approved the final manuscript.