Platelet miRNA bio-signature discriminates between dementia with Lewy bodies and Alzheimer disease: A cross-sectional study.


 Background

Dementia with Lewy bodies (DLB) is one of the most common causes of degenerative dementia after Alzheimer’s disease (AD) and presents pathological and clinical overlap with both AD and Parkinson’s disease (PD). Consequently, only one in three DLB cases is diagnosed correctly. Platelets have been previously related to neurodegeneration. They contain microRNAs (miRNAs), and their analysis may provide disease biomarkers.
Methods

Small RNAs from platelets of DLB patients and controls were analyzed by Next Generation Sequencing (NGS), and miRNAs deregulated in DLB were selected. Expression of these miRNAs was then determined and validated by LNA-based qRT-PCR in three consecutive studies from 2017 to 2019 enrolling 162 individuals, including DLB, AD, and PD patients, and healthy controls. Their predictive diagnostic potential was calculated by ROC curve analysis. miRTarbase and miRGate were used for target prediction.
Results

We profiled the whole platelet miRNA transcriptome from 7 DLB patients and 7 controls using NGS. Twenty-two selected miRNAs were further validated in independent studies (2017–2019) including DLB (n = 59), AD (n = 28), and PD (n = 24) patients, and control individuals (n = 51). Our results demonstrated downregulated expression levels for hsa-let-7d-5p (fold change 0.14; p = 0.006), hsa-miR-132-5p (0.12; p = 0.0015), hsa-miR-142-3p (0.07; p = 0.00047), hsa-miR-146a-5p (0.07; p = 0.00035), hsa-miR-150-5p (0.10; p = 0.0098), hsa-miR-25-3p (0.13; p = 0.0019) and hsa-miR-26b-5p (0.09; p = 0.0014) in DLB compared to AD. ROC curve for this seven-miRNA biosignature yielded an area under the curve (AUC) of 1. Both hsa-miR-142-3p and hsa-miR-150-5p, were down-regulated in DLB compared to controls (AUC = 0.85). Comparing AD and controls, miRNAs hsa-miR-132-5p, hsa-miR-146a-5p, hsa-miR-25-3p, and hsa-miR-6747-3p were up-regulated in AD (AUC = 0.94); and hsa-miR-128-3p and hsa-miR-139-5p were down-regulated in PD compared to controls (AUC = 0.81). Predictive target analysis identified three disease-specific pathway clusters as a result of platelet-miRNA deregulation. In DLB, pathways related to gene expression and small RNA metabolism; in AD, pathways related to stress response and RNA stress granules; and in PD pathways related to protein phosphorylation, metabolism and degradation were identified.
Conclusion

A platelet-associated bio-signature composed of 7 miRNAs is highly specific and sensitive for distinguishing DLB from AD.


Participants
The current study was conducted between 2015 and 2019. A total of 162 individuals were included from two different hospitals: Hospital Universitari Germans Trias i Pujol (Badalona, Barcelona) and Hospital Universitari de Bellvitge (L'Hospitalet de Llobregat, Barcelona). Four cohorts were recruited: DLB patients. Fifty-nine patients who ful lled criteria for probable DLB [4,21] were prospectively recruited from those visited in the Neurodegenerative disease Unit of both centres as routine clinical practice.
AD patients. Twenty-eight patients who ful lled criteria for probable AD (National Institute on Aging-Alzheimer's Association criteria, NIA-AA 2011) [22] were also consecutively recruited at the Neurodegenerative disease Unit at their routine visits, irrespective of any speci c complaint or clinical feature. AD patients were matched by age with the DLB patients.
PD patients. For comparison purposes, a group of 24 PD patients diagnosed according to the UK PD Society Brain Bank criteria [23] were included. None of these patients presented cognitive impairment, which was de ned as subjective cognitive complaints, based on the patient's and informant interview, and on the Minimental State Examination (MMSE) score, considering cognitive impairment if the MMSE punctuation was < 24 points.
In DLB, AD and PD patients, age at onset was de ned as the age when memory loss or parkinsonism was rst noticed by the patient or his/her relatives.
Control subjects. Fifty-one control individuals were selected among non-blood relatives of the patients, agematched with the DLB group.
The study was carried out in three independent phases; the rst in 2017 included 21 DLB patients and 21 controls, the second in 2018 comprised 22 DLB, and 15 AD patients, and 16 controls, and the third in 2019 contained 16 DLB, 13 AD and 24 PD patients, and 14 controls.
The study was carried out with the approval of the local Ethical Committees for Clinical Investigation of the institutions involved in the study, and a written informed consent was signed by all participants or their legal guardians according to the Declaration of Helsinki [24].
Blood collection, puri cation and characterization of platelets Peripheral blood was collected following standard procedures to minimize coagulation and platelet activation [25]. After venous puncture, blood was collected in sodium citrate tubes (BD Vacutainer®, New Jersey, USA), and processed within 2 hours following collection. After centrifugation at 500 x g for 10 minutes at room temperature to pellet red blood cells and leukocytes, the supernatant was centrifuged at 2,500 x g for 15 minutes at room temperature to obtain a platelet-rich pellet [26]. The pellet was re-suspended in 250 µL of PBS and characterized by ow cytometry for sample purity according to typical platelet size and complexity (FSC/SSC) using 100 um-Red Nile Beads (ThermoFisher) as reference and phenotypically con rmed as CD45-/CD61+ (ImmunoTools, ref21270456 and ref21330613, respectively). The analysis was performed on a FACSCanto II ow cytometer (BD).
The samples were stored at -80ºC until further processing.
Puri cation of platelet-derived small RNA Platelet-rich pellets were thawed on ice. miRNA isolation was performed using the mirVana Paris Kit (Invitrogen). Brie y, 600 µL of lysis buffer and 1/10 of miRNA Homogenate Additive Mix were added to each pellet and incubated for 10 minutes on ice after vortexing. One volume of phenol-chloroform was added, mixed and centrifuged at 10,000 x g for 5 minutes. One-third and two-thirds volume of ethanol was added in 2 consecutive steps to the miRNA containing aqueous phase and passed through a lter column. After the recommended washing steps, miRNAs were obtained with 100 µL of elution buffer and stored at -80ºC until later analysis.
MiRNA isolation from whole blood RNA isolation was carried out after collection of 3 ml of whole blood in PAXgene Blood RNA tubes (PreAnalytiX, Hombrechtikon, Switzerland) with the PAXgene Blood miRNA Kit 50, v2 (PreAnalytiX) following manufacturer's instructions. RNA quantity, purity and integrity were ascertained by the Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, USA).
Discovery Phase: miRNA sequencing and sequencing data analysis The total miRNA volume obtained from 7 DLB and 7 control samples was precipitated overnight at -20ºC with 1 µL of glycogen (20 µg/ µL), 10% 3 M AcNa (ph 4.8) and 2 volumes of ethanol. miRNAs were resuspended in 10 µL RNase free H 2 O and heated at 65ºC for 3 min. Quality control and size distribution of the puri ed small RNA was assessed by Bioanalyzer 2100 (Agilent Technologies).
Six µL of each sample were used for library preparation with the NEBNext Multiplex Small RNA Sample Preparation Set for Illumina (New England Biolabs) following the manufacturer's instructions. Individual libraries were subjected to quality analysis using a D1000 ScreenTape (TapeStation, Agilent Technologies), quanti ed by uorometry and pooled. Clustering and sequencing were performed on an Illumina Sequencer (MiSeq, Illumina, San Diego, USA) at 1 × 50c single read mode, and 200,000 reads were obtained for each sample.
Validation Phase: Reverse transcription and quantitative realtime PCR MiRNA was reverse-transcribed using the MiRCURY LNA ™ Universal cDNA Synthesis Kit II (Exiqon, Vedbaek, Denmark) according to the manufacturer's protocol. After adjusting RNA concentration to 5 ng/µL and mixing with reaction buffer and enzyme mix, a retro-transcription reaction was carried out at 42ºC for 60 minutes.
Arti cial RNA UniSp6 from the same kit was used as a retro-transcription control. Quantitative PCR (qPCR) was performed on a LightCycler 480 (Roche, Basel, Switzerland) using miRNA LNA technology and Pick&Mix PCR pre-designed panels (Exiqon) with miRNA UniSp3 as interplate calibrator control. cDNA was diluted 1:80, indications and samples were set up in duplicate.
The validation study was carried out in three independent phases, the rst in 2017, the second in 2018 and the third in 2019, including the subjects as described in 2.1.

Statistical analysis
Values for NGS data and reads are given as mean ± SD. Expression levels of the miRNAs selected for qPCR validation were determined using crossing point (Cp) values. Cp values were averaged between duplicates and normalized against UniSp6 spike-in Cp values for platelet-derived miRNA and against hsa-miR-191-5p in the case of whole blood. Relative expression changes were calculated by the -ΔΔCt method [27] and the results were further evaluated with the Wilcoxon-Mann-Whitney test [28] and the two-tailed unpaired T-test to compare the expression between two groups. When comparing more than two groups (DLB, controls, AD and PD), multiple comparisons were performed using the Kruskal-Wallis non-parametric test and Dunn's test was used for multiple corrections (GraphPad Software, Inc., La Jolla, CA, USA). In all cases, a con dence interval of 95% and a p-value below 0.05 was considered to be signi cant. To assess the diagnostic potential, the area under the ROC curve (AUC) was calculated for miRNAs with p-value < 0.01 by the Wilson/Brown method using SPSS Statistics 21 (IBM, Armonk, NY, USA) and GraphPadPrism v7 in order to determine the diagnostic sensitivity and speci city (95% C.I., AUC > 0.80 was considered as the minimum value for a useful biomarker).

miRNA target prediction and analysis
Possible targets of deregulated miRNAs (those, showing p-value < 0.01) were predicted using miRTarbase [29] accepting as target genes those that were reported only by strong evidence studies; and by miRGate [30], considering only con rmed targets. For each miRNA set, including miRNAs with expression change of miRNAs of p < 0.01, and identi ed as disease-speci c, targets from both databases were taken together, and overlapping data were removed before screening the complete list for their molecular relationship with String DataBase [31] and the Reactome online tool [32]. Gene description and most relevant information were screened also through the Uniprot database [33]. For each miRNA set, target genes were clusters by their functional characteristics, and related molecular pathway.

Results
Demographic and clinical data.
Demographic and clinical data of patients are shown in Table 1. Mean age was similar between DLB patients, AD patients and controls, according to the inclusion criteria (75.1 +/-6.8 years in the DLB group, 73.9 +/-6.7 years in the AD group; p = 0.236); however, PD patients were signi cantly younger (66.9 +/-14.9 years, p = 0.021). The male-female ratio was higher in PD and DLB, than in AD and CTRLs, but no gender-speci c expression changes were observed during data analyses. Disease duration was similar between patient groups (p = 0.068).
Platelet characterization and miRNA pro le discovery marker CD45 in our samples. Instead, we obtained a high uorescent signal for the platelet marker CD61, indicating high platelet purity and no leukocyte contamination ( Fig. 1a and Sup. Figure 2). RNA, used for the construction of NGS-libraries, showed an enriched pro le of 20-40 nucleotide molecules characteristic for small RNA and miRNA. NGS generated a mean of 1,488,787 ± 921,800 reads per sample in the control group, and 1,210,616 ± 1,868,817 reads per DLB sample. These mapped to 1,279 known different mature miRNAs, and 534 miRNAs ful lled the criteria of more than 5 reads per sample, corresponding to 430 different miRNA-precursor. Literature search revealed that 304 had been previously associated with platelets ( Fig. 1b), and 58.9% had been described in the rst platelet-miRNA pro ling studies [13,34]. Our study also con rmed let-7, miR-103 and miR-21 [15] as the most common platelet-miRNA families (Fig. 1c).
The normalized counts from NGS data were analyzed using the Wilcoxon-rank sum test, and 22 miRNAs were differentially expressed between DLB and controls, and were further validated by qPCR (hsa-miR-1343-3p, hsa-
Study II (2018). Three independent cohorts comprising newly recruited DLB patients (n = 22), AD patients (n = 15) and control subjects (n = 16) were included in the second validation study. Although 9 out of the ten miRNAs were again diminished in DLB when compared to controls, 5 out of these 9 miRNAs failed to produce signi cant results due to an elevated intra-group variability. Yet, four miRNAs con rmed the results of Study I. miRNAs hsa-miR-128-3p, hsa-miR-139-5p, hsa-miR-150-5p, hsa-miR-25-3p, were signi cantly down-regulated in DLB compared to controls, with hsa-miR-150-5p showing the most important decrease ( Table 3).
The comparison of DLB and AD miRNA expression data revealed that 9 out of the 10 miRNAs were signi cantly down-regulated in DLB compared with AD (Table 3). Only hsa-miR-150-5p was signi cantly diminished in AD when compared to controls.
The 5 miRNA sets were further studied for their usefulness as biomarkers (Fig. 3a).
ROC curve analysis ROC curves were calculated for all ve miRNA sets to assess their discrimination potential between groups.

miRNA expression in whole blood
To assess whether the results were platelet-speci c, we analyzed the 10 differentially expressed miRNAs in whole blood of DLB, PD and AD patients, and controls (n = 16, each). Only the expression of hsa-let-7d-5p and hsa-miR-132-5p was diminished in blood of PD patients in comparison with controls (Sup. miRNA target prediction Target-gene lists were obtained for the ve miRNA sets (Fig. 3a).

Discussion
In this study, we analyzed the platelet miRNA pro le in DLB patients, compared with AD and PD patients, and also with healthy controls, aiming to identify biomarkers for DLB. As a result of the rst NGS discovery phase, we selected 22 differentially expressed miRNAs which were further validated in three independent qPCR-based studies, including independent cohorts of DLB, AD, PD, and controls. Since the clinical diagnosis of DLB is still challenging, primarily because of its overlap with AD but also with PD [4], there is an urgent need for biomarkers to differentiate between these neurodegenerative disorders. Here, we de ned 3 different groups of miRNAs as speci cally deregulated in each of the three neurodegenerative disorders.
The rst group was DLB-speci c, consisted of 7 miRNAs and comprised three subsets. Hsa-miR-142-3p and hsa-miR-150-5p showed diminished expression compared to controls; these two miRNAs, together with hsa-let-7d-5p, hsa-miR-132-5p, hsa-miR-146a-5p, hsa-miR-25-3p and hsa-miR-26b-5p were down-regulated compared to AD, and hsa-miR-150-5p and hsa-miR-26b-5p were decreased in comparison with PD. Putative target genes were related to cell senescence, in ammation and signalling, and to RNA metabolism, especially to mitochondrial tRNA and gene silencing by small RNA. The disruption of RNA metabolism alterations in RNA splicing and processing, together with the deregulation of non-coding RNA has been described in several brain disorders [35]. In early PD brains, alterations in the small RNA pro le are speci cally related to tRNA fragments [36]. However, the relation between the impairment of these pathways and the development of DLB remains to be determined.
Speci cally, TIA1, required for the formation of stress granules [37], and TDRD7, a component of cytoplasmic RNA granules [38], were identi ed. TIA1 is involved in RNA splicing and post-transcriptional gene regulation, has been found in stress granules [39], and has been related with the tau oligomer and neuro brillary tangle deposition pattern in AD [37]. Stress granules are formed in the cytoplasm during transient cellular stress, and their nature and biology could be altered in neurodegenerative diseases, with chronic and long-term stress [39].
The third group of miRNAs, hsa-miR-128-3p and hsa-miR-139-5p, were signi cantly decreased in PD. Although transcriptional regulation and signal transduction were enriched in all miRNA sets, these were importantly over-represented among the target genes for these two PD-speci c miRNAs. Both RICTOR and MTOR are predicted targets for hsa-miR-128-3p and play an essential role in neuronal survival and synaptic plasticity [40]. In PD brains, MTOR expression and AKT functions are impaired [41]. MTOR over-expression impairs autophagy in genetic PD, enhancing α-synuclein deposition [42], and hsa-miR-128-3p down-regulation could be involved in the up-regulation of this pathway.
To our knowledge, the 7-miRNA biosignature composed of hsa-miR-142-3p, hsa-miR-150-5p, hsa-let-7d-5p, hsa-miR-132-5p, hsa-miR-146a-5p, hsa-miR-25-3p and hsa-miR-26b-5p, represents the rst molecular signature that permits to distinguish DLB from AD with high speci city and sensitivity. Although the precise involvement of these miRNAs in DLB pathology has yet to be clari ed, the identi cation of these biomarker candidates is particularly important, because they may improve DLB diagnosis and correspondingly, patient management, treatment and outcome. The 4th consensus report of the DLB Consortium underlined the urgent need of developing guidelines and outcome measures for clinical trials in DLB [4], this study could be crucial in the identi cation of a diagnostic biomarker to de ne inclusion/exclusion criteria for either patients with DLB, PD or AD in clinical trials.
Interestingly, three disease-speci c clusters of pathways and biological processes were identi ed as the result of platelet-miRNA deregulation. In DLB, pathways were related to gene expression and small RNA metabolism; in AD, to stress response; and in PD, to protein phosphorylation, metabolism and degradation. Since DLB and PD are synucleinopathies, the identi cation of rather similar pathways could have been expected. However, since none of the PD patients had developed dementia when samples were obtained, the involvement of different pathways may re ect the mechanisms leading to early dementia development in synucleinopathies.
The study of PD patients with dementia is needed to elucidate this question further.
No differences in miRNA expression were found in whole blood, indicating that platelet-speci c miRNA deregulation could be related to disease pathogenesis. Since platelets present neuron-like metabolic pathways, previous studies have shown that in AD, APP acts as a platelet-membrane receptor contributing mostly to soluble β-amyloid after platelet activation [43]. Mitochondrial dysfunction, higher content in phosphorylated TDP43, and morphological and structural platelet changes in AD and PD have also been reported [44,45]. Whether miRNA deregulation in platelets promotes neurodegeneration or merely re ects its effects remains to be elucidated. But a possible link between platelets and brain plasticity has been recently described [11], showing that platelets act directly on neural precursor cells in vitro, and that speci c exerciseinduced platelet activation leads to enhanced hippocampal neurogenesis [46].
Although this study has been performed in a multi-centre setting, our results need to be replicated by independent studies in other laboratories and in multi-national studies with larger cohorts. Further research should also address and con rm the alteration of the predicted biological pathways and their relationship with DLB, AD and PD. Additionally, these miRNAs should also be analyzed in groups of individuals at risk of developing a synucleinopathy or dementia, as could be individuals with idiopathic REM sleep behaviour disorder (IRBD) and mild cognitive impairment (MCI).

Conclusion
In summary, two main ndings must be highlighted. First, we have shown that the miRNA content from platelets may represent a promising source of biomarkers. In particular, we de ned a 7-miRNA bio-signature that may represent a useful biomarker for the differentiation between DLB and AD patients. Second, we de ned speci c clusters of pathways and biological processes for DLB, AD and PD, underlining that the development of the different diseases is, at least in part, platelet-driven by affecting speci c pathways.         were associated with DLB in comparison to AD and PD.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.