A 9-microRNA Signature in Serum Serves as a Noninvasive Biomarker in Early Diagnosis of Alzheimer’s Disease

Abstract

Alzheimer’s disease (AD) is the most common type of age-related neurodegenerative disorder; nevertheless, nowadays there are no reliable biomarkers or non-invasive techniques available for its early detection. Recent studies have indicated that the circulating level profiles of microRNAs (miRNAs) have the potential to be used as valuable biomarkers for diagnosis, staging, and progress monitoring of various diseases. Here we report a novel 9-miRNA signature (hsa-miR-26a-5p, hsa-miR-181c-3p, hsa-miR-126-5p, hsa-miR-22-3p, hsa-miR-148b-5p, hsa-miR-106b-3p, hsa-miR-6119-5p, hsa-miR-1246, and hsa-miR-660-5p) that can be utilized as biomarker for detecting AD. We respectively profiled the serum miRNAs from 19 AD patients and 9 healthy control (HC) participants using the Next-Generation Sequencing (NGS). The NGS results were validated by quantitative real-time polymerase chain reaction (qRT-PCR) on a larger cohort of 121 AD and 86 HC cases. All the patients were divided into three groups (mild, moderate, and severe AD) based on the Mini-Mental State Examination (MMSE) and Clinical Dementia Rating (CDR). Our research indicates that abnormal expression of distinct serum miRNAs occurs at different stages of AD. The difference of the area under the receiver operator characteristics curve (AUC) between the AD and the HC is between 70% and 85%. Among the 9 miRNAs, hsa-miR-22-3p has the best sensitivity (81.8%) and specificity (70.9%). The miRNA-panel is more valuable for AD diagnosis. The data suggest that the differentially expressed serum miRNAs could be used as biomarkers to improve the diagnosis of AD, particularly at the early stage, and to classify its clinical stages.

Keywords

Alzheimer’s disease biomarker microRNAs next-generation sequencing quantitative Real-Time PCR serum

INTRODUCTION

Alzheimer’s disease (AD) is the most common form of dementia characterized by the accumulation of amyloid-β (Aβ) plaques and the formation of neurofibrillary tangle (NFT) [1]. The definitive diagnosis of AD can only be confirmed by autopsy, since the neurodegenerative changes are believed to begin years before the clinical presentation of dementia. The clinical diagnosis of AD is based on medical history, clinical examination, neuropsychological testing, imaging examination, and laboratory assessments. However, these methods have not yet proved sensitive and specific enough for the definitive diagnosis of AD, especially at the early stages of the disease. Although Aβ neuroimaging using positron emission tomography (PET) and the measurement of biomarkers in the cerebrospinal fluid (CSF) have shown excellent diagnostic accuracy for AD, the high costs and invasiveness of these methods, respectively, restrict their application [2 –4]. Consequently, there is a crucial need for reliable biomarkers for the early detection of AD in order to implement preventative treatment strategies.

MicroRNAs (miRNAs) are a class of endogenous small non-coding RNAs with a length of 21–25 nucleotides that regulate gene expression at the post-transcriptional level by imperfect complementary sequence binding to the 3^′ untranslated region (3^′UTR) of the target mRNAs, leading to their translational inhibition and degradation [5 –7]. Accumulating evidence suggest that alterations in specific miRNAs levels are associated with various disorders and therefore these miRNAs could be used as biomarkers for the diagnosis of various diseases such as cancer and cardiovascular diseases [8 –12]. Previous studies have found that miRNAs, abundantly expressed in the central nervous system, showed a high degree of temporal and spatial specificity and were mainly involved in neuronal formation, differentiation, and synaptic plasticity[5 , 13–15]. Furthermore, some studies indicated that the microvesicles containing miRNAs can be released into peripheral blood through the blood-brain barrier [16 –18]. Importantly, many researchers have reported that changes of miRNAs levels can be detected in the serum of AD patients [19, 20]. Serum is an appealing source of biomarkers due to its easy collection with minimal discomfort to the patient, encouraging greater compliance in clinical trials and frequent testing. Therefore, serum miRNAs have the potential to be used as biomarkers for the diagnosis of AD.

Recently, with the development of high-throughput next-generation sequencing (NGS), researchers are able to analyze miRNAs of biological fluids in a genome-wide scale for the discovery of disease biomarkers [15, 21]. However, there were no NGS studies on miRNAs screens for biomarkers to monitor the progression of AD, particularly at its early stages. Therefore, the primary aim of this study is to investigate whether serum miRNAs of AD patients can be used as diagnostic biomarkers, particularly in its presymptomatic and early stage, and monitor the progression of disease. We obtained the miRNA profiles of human serum by using the NGS approach from three groups of AD patients (mild, moderate, and severe AD) and healthy controls (HC). After validation of qRT-PCR results in the larger cohort of participants, 9 miRNAs that showed an association with the clinical stages of AD were relevant to use as biomarkers for the early diagnosis of AD.

MATERIALS AND METHODS

Participants

The participants were selected from the out-patient clinic at the Shandong Provincial Hospital and the Shandong Mental Health Center, between March 13, 2014, and May 6, 2016. A written informed consent was obtained from all the participants. Ethical approval for this study was obtained by the institutional ethics committees of the Shandong Provincial Hospital and the Shandong Mental Health Center. The study was performed in accordance with the Helsinki declaration.

The clinical diagnosis of AD fulfilled the criteria set by the National Institute of Neurological and the Communicative Diseases and Stroke–Alzheimer’s Disease and Related Disorders Association (NINCDS–ADRDA) [22]. In addition, all the participants were tested by general laboratory examinations and neurological examinations in order to exclude other common geriatric diseases and neurological disorders. The degree of cognitive impairment was assessed by the Mini-Mental State Examination (MMSE) and the Clinical Dementia Rating (CDR) questionnaire. According to their MMSE and CDR score, we divided the participants into four groups: healthy controls (MMSE score: 27–30, CDR-0), mild AD (MMSE score: 20–26, CDR-0.5 or 1), moderate AD (MMSE score: 10–19, CDR-2), and severe AD (MMSE score: 0–9, CDR-3) [23]. In the primary screen, individual serum samples from 28 participants (9 controls; 6 mild, 7 moderate, and 6 severe AD cases) used in NGS were selected randomly, and they were a small fraction of the total samples. For validation purposes, we analyzed the expression of single miRNAs using qRT-PCR in the same samples as used for NGS, if sufficient amounts of RNA were available. We further expanded the number of samples, resulting in a total of 207 samples (86 controls; 31 mild, 52 moderate, and 38 severe AD cases) analyzed by qRT-PCR. This is similar to the method the used in Leidenger’s article [23].

Serum collection

Fasting blood samples (3 ml per subject) were collected in BD Vacutainer SST tubes (BD, New Jersey, NJ, EEUU). The serum was collected within 2 h after blood was isolated from patients, avoiding any wild swings, and the serum samples hemolyzed visually were removed. The tubes were kept in vertical position for 30 min and then centrifuged at 3,000 rpm for 10 min at room temperature, followed by centrifugation at 16,000 rpm for 5 min at 4C. All remaining serum samples were taken to measure the oxyhemoglobin absorbance using Nanodrop 2000 spectrophotometer (Thermo). Hemolyzed samples showing a peak at 414 nm were removed [24]. In addition, the relative expression of miR-451 and miR-23a were also used for indication of hemolysis. Accordingly, a serum sample with ΔCt (miR-23a- miR-451)>5 shows hemolysis, and was not used in the study [24, 25]. Other serum samples were stored at –80C for further analysis. Repeated freeze/thaw cycles were avoided. We collected a total of 207 serum samples from 121 AD and 86 HC cases that matched the age and gender criteria.

Screening of miRNA biomarkers by next-generation sequencing

Serum samples from 28 participants (9 controls; 6 mild, 7 moderate, and 6 severe AD cases) were selected for NGS. Total RNA was extracted respectively using Trizol reagent (Invitrogen, CA, USA), following the manufacturer’s instructions. Total RNA quantity and purity analysis were performed on the Bioanalyzer 2100, by using the RNA 6000 Nano LabChip Kit (Agilent, CA, USA), with RIN number >7.0. Approximately 1 μg of total RNA from each sample were used to prepare the small RNA library, according to the instructions of the TruSeq Small RNA Sample Prep Kits (Illumina, San Diego, USA). Next, we performed the single-end sequencing (36 bp) on an Illumina Hiseq 2500 at the LC-BIO (Hangzhou, China) following the vendor’sinstructions.

The data processing provided by LC Sciences Service (Houston, TX; http://www.lcsciences.com). The raw reads were subjected to the Illumina pipeline filter (Solexa 0.3), and then the dataset was further processed with an in-house program, ACGT101-miR(LC Sciences, Houston, TX, USA) to remove adapter dimers, junk, low complexity, common RNA families (rRNA, tRNA, snRNA, snoRNA) and repeats. Subsequently, unique sequences with length in 18∼26 nucleotide were mapped to specific species precursors in miRBase 20.0 (http://www.mirbase.org/pub/mirbase/CURRENT/) by BLAST search to identify known miRNAs and novel 3p- and 5p- derived miRNAs. Length variation at both 3’ and 5’ ends and one mismatch inside of the sequence were allowed in the alignment. The unique sequences mapping to specific species mature miRNAs in hairpin arms were identified as known miRNAs. The unique sequences mapping to the other arm of known specific species precursor hairpin opposite to the annotated mature miRNA-containing arm were considered to be novel 5p- or 3p- derived miRNA candidates. The remaining sequences were mapped to other selected species precursors (with the exclusion of specific species) in miRBase 20.0 by BLAST search, and the mapped pre-miRNAs were further BLASTed against the specific species genomes to determine their genomic locations. The above two we defined as known miRNAs. The unmapped sequences were BLASTed against the specific genomes, and the hairpin RNA structures containing sequences were predicated from the flank 80 nt sequences using RNAfold software (http://rna.tbi.univie.ac.at/cgi-bin/RNAfold.cgi). The database of Genome (http://ftp.ensembl.org/pub/release-76/fasta/homo sapiens/dna/) and the database of mRNA (http://ftp.ensembl.org/pub/release-76/fasta/homo sapiens/cdna/) was used. The miRNAs expression assays and the statistical analysis of the sequencing data were performed by the LC Sciences.

Validation of miRNA biomarkers by qRT-PCR

The differentially expressed candidate miRNAs were validated using qRT-PCR on a larger cohort of serum samples. Samples were also divided into four groups (healthy controls; mild, moderate, and severe AD). Up to date, there are no commonly approved endogenous control microRNAs in the serum that can be used for normalization. We have chosen hsa-miR-451 as the endogenous control, owing to its stable expression across most of AD samples, and we added the same amount of cel-miR-39 (Takara, Japan) to an equal volume of serum as the heterogeneous control [18]. Small RNAs were extracted from 200 μl of serum sample aliquots from each participant, using the miRCURY™ RNA Isolation Kit–Biofluids (Exiqon, Vedbaek, Denmark), following the manufacturer’s procedure. The kit can be used for the isolation and purification of all RNAs smaller than 1000 nt, including mRNA, tRNA, microRNA, and small interfering RNA (siRNA). The small RNAs from each sample preparation were finally eluted in 25 μl of RNase-free water. The RNA Spike-in Kit (Exiqon, Vedbaek, Denmark) was used to monitor the yield and quality of small RNAs in the eluate.

The equal volumes (10 μl) of small RNAs extracted from serum was polyadenylated by poly (A) polymerase and then reverse transcribed to cDNA by using universal adaptor primer. qRT-PCR was performed with the SYBR^® Prime ScriptTM miRNA RT-PCR Kit (Takara) in a total volume of 20 μl per reaction containing 1 μl cDNA, and used the LightCycler^® 480 (Roche) according to the manufacturer’s instructions. All miRNA specific forward primers were purchased from Takara. The Uni-miR qPCR Primer from the kit was used as reverse primer. RNase-free water was used as negative control. Cycling conditions for real-time PCR were 95°C for 5 min, 45 cycles of 95°C for 10 s, 60°C for 15 s, and 72°C for 15 s followed by a melt-curve analysis to evaluate PCR specificity. All the samples were measured in triplicate and the calculated mean value was used for further analysis. The ΔCt was calculated by subtracting the mean Ct values of hsa-miR-451 and cel-miR-39 from the Ct values of the miRNA of interest. The Δ ΔCt was then calculated by subtracting ΔCt of the control from ΔCt of disease. The fold change of deregulated miRNA was calculated by the equation 2^-ΔΔCt.

Statistical analysis

In the screening by NGS, the normalization method is as follows: 1) Find a common set of sequences among all samples; 2) Construct a reference data set. Each data in the reference set is the copy number median value of a corresponding common sequence of all samples; 3) Perform 2- based logarithm transformation on copy numbers (log₂ (copy#)) of all samples and reference data set; 4) Calculate the log₂ (copy#) difference (Δlog₂ (copy#)) between individual sample and the reference data set; 5) Form a subset of sequences by selecting |Δlog₂ (copy#)| <2, which means less than (2² =) 4 fold change from the reference set; 6) Perform linear regressions between individual samples and the reference set on the subset sequences to derive linear equations y = a _i x+b _i where a _i and b _i are the slop and interception, respectively, of the derived line, x is log₂ (copy#) of the reference set, and y is the expected log₂ (copy#) of sample i on a corresponding sequence; 7) Calculate the mid value x _mid = (max (x)- min (x))/ 2 of the reference set. Calculate the corresponding expected log₂ (copy#) of sample i, y _i,mid = a _i x _mid +b _i. Let y _r,mid = x _mid. Let Δ y _i = y _r,mid- y _i,mid, which is the logarithmic correction factor of sample i. We then derive the arithmetic correction factor f _i = 2 ^Δyi of sample i; and 8) Correct copy numbers of individual samples by multiplying corresponding arithmetic correction factor f _i to original copy numbers.

Age and MMSE score of groups were compared using ANOVA with Bonferroni’s post-hoc test for multiple comparison. Gender differences between groups were determined using Chi-square test. The miRNA expression levels of HC and AD were compared using Student’s t test for single comparison. The miRNA expression levels of four groups (HC, mild AD, moderate AD, and severe AD) were compared using ANOVA with Bonferroni’s post-hoc test for multiple comparison. The area under the receiver operating characteristic (ROC) curve (AUC) was used to evaluate the diagnostic performance of each miRNA. The optimal cut-off points were determined based on the maximized sum of sensitivity and specificity and the binary logistic regression model was used to establish the ROC curve of combined miRNAs. The correlations between miRNA expression levels and MMSE score were assessed with the Pearson’s correlation coefficient. The statistical power for the sample sizes used, particularly for the analysis of subsamples, is calculated by using G*Power 3.1. SPSS statistics 19.0 (IBM, New York, NY) and Prism5.0 (GraphPad, La Jolla, CA) were used to perform statistical analyses. Group differences were considered significant when p < 0.05.

RESULTS

miRNA discovery

All the participants in this study underwent several examinations, such as general biochemical tests, neurological examinations, imaging tests, and cognitive impairment evaluation. The demographic and clinical makeup details are summarized in Table 1. To identify any potential AD-related miRNA biomarkers we examined the serum from 28 participants(9 controls; 6 mild, 7 moderate, and 6 severe AD cases) using NGS.

Table 1

Demographics and clinical makeup of participants in the study

	HC	mild AD	moderate AD	severe AD	p-value
Next-Generation Sequencing (NGS) set
n	9	6	7	6
Age (y) (mean±s.d.)	70.81 (8.56)	71.02 (6.51)	69.17 (7.20)	75.23 (7.52)	0.944
Gender (male:female)	5:4	4:2	3:4	3:3	0.854
MMSE (mean±s.d.)	29.40 (0.83)	23.21 (1.67)	15.83 (2.73)	6.33 (3.58)	<0.001
CDR	0	0.5 or 1	2	3
RT-PCR set
n	86	31	52	38
Age (y) (mean±s.d.)	67.95 (5.76)	71.43 (6.21)	72.52 (7.45)	74.12 (5.25)	0.907
Gender (male:female)	45:41	16:15	28:24	17:21	0.843
MMSE (mean±s.d.)	28.91 (0.94)	24.06 (1.34)	15.51 (2.56)	7.01 (3.73)	<0.001
CDR	0	0.5 or 1	2	3

HC, healthy control; AD, Alzheimer’s disease; MMSE, Mini-Mental State Examination; CDR, Clinical Dementia Rating.

NGS generated a mean total of 11,592,444 raw reads which were used to construct the sRNA libraries of the HC and the AD samples. The valid reads of the HC and the AD libraries were obtained after removal of the corrupted adapter sequences, reads with length <18 and >26 nt, and junk reads (Supplementary Tables 1 and 2). The majority of valid sRNAs in both libraries were of 18–26 nt in length, which is within the typical size range for RNA processing products (Supplementary Figure 1).The NGS generated sequences were mapped to small RNA sequences in the miRBase 20.0 database. The analysis not only identified confirmed miRNAs, but also many novel miRNAs. Overall, 3,323 pre-miRNAs, corresponding to 3,387 mature miRNAs, were detected in our study. All the relevant information of the detected miRNAs is summarized in theSupplementary Table 3.

The normalized NGS counts AD and HC were analyzed by Student t test (Supplementary Table 4). The significance threshold was set at p≤0.05 for each test. Differential expression of a particular miRNA in a sample was defined when the sample contained at least 50 copies and when a±1.8 fold difference was observed between the AD versus the HC. A total of 32 miRNAs were identified that were differentially expressed in AD (relative to the HC). The miRNA expression levels of four groups (HC, mild AD, moderate AD, and severe AD) were compared using ANOVA with Bonferroni’s post-hoc test for multiple comparison (Supplementary Table 5). Base on the results of ANOVA, we selected a further 13 miRNAs with altered expression (Table 2), which were differential expression in comparison between all groups (Table 3 and Fig. 1). Among the 13 miRNAs, 6 miRNAs (hsa-miR-26a-5p, hsa-miR-181c-3p, hsa-miR-584, hsa-miR-126-5p, hsa-miR-22-3p, hsa-miR-148b-5p) were down-regulated and 7 miRNA (hsa-miR-221-3p, hsa-miR-106b-3p, hsa-miRNA-144-5p, hsa-miR-6119-5p, hsa-miR-1388-3p, hsa-miR-1246, hsa-miR-660-5p) were upregulated. After Bonferroni’s post-hoc test, we found the 13 miRNAs have obvious difference between AD subgroups and HC, but they exhibited no difference between AD subgroups (Table 3). Calculation of the statistical power for the sample sizes used, particularly for the analysis of subsamples, was carried out. The results of the statistical power were exhibited in the Tables 2 and 3.

Table 2

The 13 miRNA biomarkers screened by next-generation sequencing

miR name	Sequences	miRNA reads		up/down	Fold change	p-value	Statistical
		(mean)					power
		HC	AD
hsa-miR-26a-5p	UUCAAGUAAUCCAGGAUAGGCU	56,021	28,430	Down	0.51	0.035	0.87
hsa-miR-181c-3p	ACCAUCGACCGUGAGUGGACC	254	94	Down	0.37	0.049	0.86
hsa-miR-584	UUAUGGUUUGCCUGGGACUGA	4,258	1,544	Down	0.36	0.021	0.93
hsa-miR-126-5p	CAUUAUUACUUUUGGUACGCG	45,191	20,974	Down	0.46	0.033	0.90
hsa-miR-22-3p	AAGCUGCCAGUUGAAGAACUGU	119,009	63,473	Down	0.53	<0.001	0.98
hsa-miR-148b-5p	GAAGUUCUGUUAUACACUCAGGCU	307	146	Down	0.48	0.040	0.78
hsa-miR-221-3p	AGCUACAUUGUCUGCUGGGUU	28,670	50,134	Up	1.75	0.001	0.87
hsa-miR-106b-3p	CCGCACUGUGGGUACUUGCU	3,950	7,443	Up	1.88	<0.001	0.85
hsa-miR-144-5p	GGAUAUCAUCAUAUACUGUAAG	3,467	6,283	Up	1.81	0.004	0.79
hsa-miR-6119-5p	AGAGGUAAAAAAUUGAUUUGAC	16	103	Up	6.39	0.001	0.77
hsa-miR-1388-3p	AUCUCAGGUUCGUCAGCCCGC	22	82	Up	3.66	0.014	0.57
hsa-miR-1246	AAUGGAUUUUUGGAGCAGGGAGU	8	57	Up	7.44	0.018	0.51
hsa-miR-660-5p	UACCCAUUGCAUAUCGGAGCUG	103	360	Up	3.48	0.013	0.58

HC, healthy control; AD, Alzheimer’s disease.

Table 3

The differential expression analysis of 13 miRNAs screened by NGS between mild, moderate, severe AD and healthy control

miR name	miRNA reads (mean)				p-value	Statistical
	HC	mild AD	moderate AD	severe AD	ANOVA	power
hsa-miR-26a-5p	56,021	27,128*	28,140*	30,071	0.032	0.58
hsa-miR-181c-3p	254	80*	111	87	0.034	0.58
hsa-miR-584	4,258	1,396*	1,796*	1,398*	0.007	0.71
hsa-miR-126-5p	45,191	18,510*	19,754*	24,861	0.017	0.65
hsa-miR-22-3p	119,009	54,842*	58,390*	78,035*	<0.001	0.87
hsa-miR-148b-5p	307	155	135	150	0.097	0.46
hsa-miR-221-3p	28,670	56,130*	45,491	49,556	0.016	0.65
hsa-miR-106b-3p	3,950	8,625*	6,929	6,861	0.021	0.63
hsa-miR-144-5p	3,467	7,077*	5,279	6,662	0.038	0.57
hsa-miR-6119-5p	16	87	109	111	0.093	0.46
hsa-miR-1388-3p	22	96	37	119*	0.056	0.53
hsa-miR-1246	8	50	9	119*	0.006	0.72
hsa-miR-660-5p	103	355	188	565*	0.044	0.55

The results of Bonferroni’s post-hoc test: *Different from HC, p < 0.1.

Fig.1

Heatmap of differentially expressed miRNAs obtained from AD patients and healthy controls. The hierarchical clustering of dysregulated miRNAs was performed by using the TMEV software. We selected 13 miRNAs with altered expression, which were differential expression in comparison between all groups. The first dendrogram node contains 6 miRNAs that were found to be downregulated (hsa-miR-26a-5p, hsa-miR-181c-3p, hsa-miR-584, hsa-miR-126-5p, hsa-miR-22-3p, hsa-miR-148b-5p). The second dendrogram node contains 7 miRNAs that were found to be upregulated (hsa-miR-221-3p, hsa-miR-106b-3p, hsa-miR-144-5p, hsa-miR-6119-5p, hsa-miR-1388-3p, hsa-miR-1246, hsa-miR-660-5p). Participants samples included Group2 (healthy controls), Group4 (mild AD), Group5 (moderate AD) and Group6 (severe AD).

Validation of miRNA expression

The 13 candidate miRNAs with differential expression were validated by qRT-PCR in the serum of 207 participants (86 controls; 31 mild, 52 moderate, and 38 severe AD cases). The demographic and clinical makeup details are presented in Table 1. miRNA specific forward primers are listed in Table 4. As previously mentioned, ΔCt (miR-23a- miR-451) values of all serum samples were smaller than 5, which shows that these samples were not hemolyzed. Except the hsa-miR-1388-3p, Ct values for all candidate miRNAs were <35 in the majority of samples and demonstrated a single melt peak in a preliminary profiling analysis. The hsa-miR-584 and hsa-miR-221-3p expression levels were not statistically significant different between the AD versus the HC. Therefore, these three miRNAs were eliminated from our candidate miRNAs. Comparison of the expression analysis results (fold changes) obtained by NGS and qRT-PCR (Fig. 2), revealed that 9 miRNAs had a similar differential expression pattern. However, the hsa-miR-144-5p was upregulated in the NGS screen and downregulated according to qRT-PCR validation. The reason of this discrepancy might be attributed to the differential sample cohorts used by the two approaches. Calculation of the statistical power for the sample sizes used, particularly for the analysis of subsamples, was carried out. The values of the statistical power for the sample sizes were all close to 1 in RT-PCR validation of large samples.

Table 4

List of primer sequences used in the present study

Primer	Sequence
hsa-miR-26a-5p	5’-TTCAAGTAATCCAGGATAGGCT-3’
hsa-miR-181c-3p	5’-ACCATCGACCGTGAGTGGACC-3’
hsa-miR-584	5’-TTATGGTTTGCCTGGGACTGA-3’
hsa-miR-126-5p	5’-CATTATTACTTTTGGTACGCG-3’
hsa-miR-22-3p	5’-AAGCTGCCAGTTGAAGAACTGT-3’
hsa-miR-148b-5p	5’-GAAGTTCTGTTATACACTCAGGCT-3’
hsa-miR-221-3p	5’-AGCTACATTGTCTGCTGGGTT-3’
hsa-miR-106b-3p	5’-CCGCACTGTGGGTACTTGCT-3’
hsa-miR-144-5p	5’-GGATATCATCATATACTGTAAG-3’
hsa-miR-6119-5p	5’-AGAGGTAAAAAATTGATTTGAC-3’
hsa-miR-1388-3p	5’-ATCTCAGGTTCGTCAGCCCGC-3’
hsa-miR-1246	5’-AATGGATTTTTGGAGCAGGGAGT-3’
hsa-miR-660-5p	5’-TACCCATTGCATATCGGAGCTG-3’
hsa-mir-451	5’-AAACCGTTACCATTACTGAGTT-3’
hsa-miR-23a	5’-ATCACATTGCCAGGGATTTCC-3’
cel-miR-39	5’-TCACCGGGTGTAAATCAGCTTG-3’

Fig.2

Comparison of the expression analysis results (fold changes) obtained by NGS and qRT-PCR between the AD patients versus the healthy controls. 9 miRNAs had a similar differential expression pattern. However, the hsa-miR-144-5p was upregulated in the NGS screen and downregulated according to the qRT-PCR validation. The p-values were determined by the Student’s t-test.

The ROC curve analysis highlighted remaining 9-miRNA signatures as potential biomarkers for AD diagnosis (Fig. 3A-I). Computation of the area under the ROC curve (AUC), showed a high AUC value, between 70.0% and 85.3%, indicating that each of these miRNAs has sufficient power to differentiate between AD and HC. Among the 9 miRNAs, hsa-miR-22-3p has the best sensitivity (81.8%) and specificity (70.9%). In addition, such analysis of ROC curve also showed that the combination of hsa-miR-26a-5p/ hsa-miR-181c-3p/ hsa-miR-22-3p/ hsa-miR-148b-5p/ hsa-miR-106b-3p/ hsa-miR-6119-5p/ hsa-miR-660-5p (miR-panel) conferred high relevance for AD diagnosis (Fig. 3J). The miR-panel showed higher accuracy of AUC (98.6%), higher sensitivity (81.8%), and higher specificity (70.9%). Results of ROC curve analysis are shown in Table 5. In addition, we analyzed the potential correlation between the levels of 9 serum miRNAs and MMSE score. The results of Pearson correlation analysis showed that 5 miRNAs (hsa-miR-26a-5p, hsa-miR-181c-3p, hsa-miR-126-5p, hsa-miR-22-3p, hsa-miR-148b-5p) were positively correlated to MMSE score and 4 miRNAs (hsa-miR-106b-3p, hsa-miR-6119-5p, hsa-miR-1246, hsa-miR-660-5p) were negatively correlated to MMSE score (Fig. 4).

Fig.3

Receiver-operating characteristic (ROC) curve analysis results for the differentiation between the AD patients and the healthy controls. The figure shows the ROC curves of individual miRNAs in the 9-miRNA signature (A-I) and the miR-panel (J). The areas under the ROC curve (AUC) are between 70.0% and 85.3%, indicating that each of these miRNAs has sufficient power to differentiate between the AD patients and the healthy controls. The AUC of the miR-panel is 98.6%, indicating that the miR-panel is more valuable for AD diagnosis.

Table 5

The ROC results of the candidate miRNAs and the miRNA panel

miR name	p-value	AUC (95% CI)	Sensitivity (%)	Specificity (%)	Cut-off	Youden’s index
hsa-miR-26a-5p	<0.05	0.754 (0.686 to 0.821)	85.1	57.0	0.2095	0.421
hsa-miR-181c-3p	<0.05	0.784 (0.720 to 0.847)	71.9	73.3	0.0022	0.452
hsa-miR-126-5p	<0.05	0.717 (0.645 to 0.788)	72.7	60.5	0.3082	0.332
hsa-miR-22-3p	<0.05	0.853 (0.803 to 0.904)	81.8	70.9	0.1131	0.528
hsa-miR-148b-5p	<0.05	0.724 (0.654 to 0.794)	64.5	72.1	0.0018	0.366
hsa-miR-106b-3p	<0.05	0.828 (0.772 to 0.885)	76.0	75.6	0.0283	0.516
hsa-miR-6119-5p	<0.05	0.790 (0.729 to 0.851)	75.2	69.8	0.0002	0.450
hsa-miR-1246	<0.05	0.768 (0.706 to 0.831)	52.1	90.7	0.0004	0.428
hsa-miR-660-5p	<0.05	0.700 (0.630 to 0.770)	49.6	82.6	0.0024	0.322
miRNA-panel	<0.05	0.987 (0.976 to 0.998)	93.4	98.8	0.7005	0.922

AUC, area under curve; CI, confidence interval.

Fig.4

Abnormal expression levels of the 9 miRNAs correlate with the MMSE score. The positive correlations between the levels of miRNAs and MMSE score: hsa-miR-26a-5p (r = 0.196, p = 0.005), hsa-miR-181c-3p (r = 0.412, p < 0.001), hsa-miR-126-5p (r = 0.280, p < 0.001), hsa-miR-22-3p (r = 0.454, p < 0.001), hsa-miR-148b-5p (r = 0.332, p < 0.001); The negative correlations between the levels of miRNAs and MMSE score: hsa-miR-106b-3p (r = –0.463, p < 0.001), hsa-miR-6119-5p (r = –0.530, p < 0.001), hsa-miR-1246 (r = 0.–678, p < 0.001), hsa-miR-660-5p (r = –0.551, p < 0.001). All values were normalized to the endogenous hsa-miR-451 and the external cel-miR-39 levels, and are represented as log10 scales at the Y-axis.

The miRNA expression levels of four groups (HC, mild AD, moderate AD, and severe AD) were compared using ANOVA with Bonferroni’s post-hoc test for multiple comparison. The results of ANOVA showed that the 9 miRNAs were all differential expression in comparison between all groups, and the results of Bonferroni’s post-hoc test showed that the 9 miRNAs have obvious difference between AD subgroups and HC (Table 6 and Fig. 5). 6 miRNAs (hsa-miR-26a-5p, hsa-miR-181c-3p, hsa-miR-126-5p, hsa-miR-22-3p, hsa-miR-148b-5p, and hsa-miR-106b-3p) out of 9 were differentially expressed as “mild” stage of AD. 2 miRNAs (hsa-miR-6119-5p and hsa-miR-1246) were differentially expressed at the “moderate” stage, and thebta-miRNA-660 was differentially expressed at the “severe” AD stage, while these 9-miRNAs showed biological accuracy in the classification of AD stages. In addition, 4 miRNAs have difference between the AD subgroups. As follows: the expression of hsa-miR-26a-5p and hsa-miR-6119-5p showed significant differences in mild AD than moderate or severe stages. Moreover,expression of hsa-miR-1246 and hsa-miR-660-5p in severe AD were different in mild and moderate stages (Table 6). So the 9 miRNAs have the ability for improvement of AD diagnosis, for classification of patients to a specific subgroup. But the differences of miRNAs between the subgroups are insufficient to be used as progression biomarkers to evaluate the disease development.

Table 6

The differential expression analysis of 9 miRNAs validated by qRT-PCR between mild, moderate, severe AD and healthy control

miR name	2^-ΔCt (mean)				p-value
	HC	mild AD	moderate AD	severe AD	ANOVA
hsa-miR-26a-5p	2.12E-01	1.49E-01*	1.66E-01*	1.83E-01*^#	<0.001
hsa-miR-181c-3p	2.59E-03	2.09E-03*	1.82E-03*	2.01E-03*	<0.001
hsa-miR-126-5p	3.36E-01	2.59E-01*	2.82E-01*	2.72E-01*	<0.001
hsa-miR-22-3p	2.37E-01	6.73E-02*	7.96E-02*	5.63E-02*	<0.001
hsa-miR-148b-5p	4.98E-03	1.79E-03*	1.39E-03*	1.57E-03*	<0.001
hsa-miR-106b-3p	2.30E-02	6.23E-02*	6.44E-02*	6.63E-02*	<0.001
hsa-miR-6119-5p	1.62E-04	2.10E-04	4.84E-04*^#	4.61E-04*^#	<0.001
hsa-miR-1246	2.85E-04	3.03E-04	3.50E-04*	5.99E-04*^#&	<0.001
hsa-miR-660-5p	2.04E-03	2.21E-03	2.23E-03	3.37E-03*^#&	<0.001

The results of Bonferroni’s post-hoc test: *Different from HC, p < 0.05; ^#Different from mild AD, p < 0.05; ^&Different from moderate AD, p < 0.05. HC, healthy control; AD, Alzheimer’s disease.

Fig.5

Scatter plots of the validated miRNAs differentially expressed in comparison between all groups. The levels of the 9 differentially expressed miRNAs in the serum from 207 participants (86 controls; 31 mild, 52 moderate, and 38 severe AD cases) were measured by qRT-PCR. Each point represents the mean value calculated from the triplicate results. All values were normalized to the endogenous hsa-miR-451 and the external cel-miR-39 levels, and are represented as log10 scales at the Y-axis. The horizontal line at each group represents the median value. The p-values were determined by Bonferroni’s post-hoc test.

DISCUSSION

Nowadays, there are no available biomarkers sensitive and specific enough to be used for the definitive diagnosis of AD, particularly at the early stages of the disease. Circulating serum miRNAs can offer a rapid, non-invasive and cost effective method for the diagnosis of many diseases [26 –28].In the current study, we report 9 serum miRNAs (hsa-miR-26a-5p, hsa-miR-181c-3p, hsa-miR-126-5p, hsa-miR-22-3p, hsa-miR-148b-5p, hsa-miR-106b-3p, hsa-miR-6119-5p, hsa-miR-1246 and the hsa-miR-660-5p) derived from genome-wide miRNA expression profiling of serum samples. These miRNAs can be utilized as biomarkers for AD detection. In addition, we established a miR-panel (combination of hsa-miR-26a-5p/ hsa-miR-181c-3p/ hsa-miR-22-3p/ hsa-miR-148b-5p/ hsa-miR-106b-3p/ hsa-miR-6119-5p / hsa-miR-660-5p) by the Binary logistic regression to diagnose AD, and the miR-panel is more valuable for AD diagnosis. A possible future direction could be building a miRNA library from serum samples of AD patients, and combining these serum miRNA signatures with other clinical detection methods to establish a more comprehensive patient’s diagnosis and treatment strategy; should these novel findings help establishing new AD prognosis strategies with increased efficacy.

In the current study, we focused on serum samples to detect differentially expressed miRNAs in AD. Today, high-throughput techniques such as genome-wide screening method is performed to determine miRNA levels in the biological fluids [15, 21]. Although some miRNAs from blood samples of AD patients were previously reported as potential biomarkers by using NGS (validation by qRT-PCR), the detect of these markers remained limited and our findings do not overlap with these markers. Out of 9 markers that we identified, only hsa-miR-26a-5p were reported by Leidinger et al. [23]. Thisdiscrepancy may likely result from common variations in the chosen libraries. For example, miRNA extracts were derived from different samples such as, blood, plasma, serum, and the exosomes of serum. In addition, sample size differed in the screening of NGS and was often not large enough to conclude a full analysis. miRNAs are a class of endogenous transcription regulatory factors that vary with the clinical progress of AD. Serum sample libraries from different studies have different patient cohort size from different stages of the disease, and this can be another factor in the differences between miRNA profiling studies. Thus, it is important to carefully group subjects according to AD severity when researching miRNAs as biomarkers. Another factor to consider is the ethnic (and thereby genomic) differences between study cohorts. Experimental factors such as data normalization can also account for differences.

Our research also indicates that the abnormal expression of these distinct serum miRNAs occurs at different stages of AD. Although it was known that alterations in the miRNAs network could increase the risk of AD, we have been not sure how the expression level profile of some specific miRNAs varies with the clinical progress of the AD. To date, limited number of studies attempted to measure the expression profile of circulating miRNAs in different stages of AD. Cheng et al. identified an AD-specific 16-miRNA signature, due to the limitation of the sample size, and failed to distinguish mild cognitive impairment from AD [18]. Leidinger reported detection of a 12-miRNA blood-based signature, but showed no significant expression differences between mild, moderate and severe AD stages, based on the MMSE scoring system [23]. In order to find AD progression-dependent serum miRNAs, we also divided the AD patients into mild, moderate and severe AD stages. We detected 9 novel circulating miRNAs with altered expression, which are closely associated with the classification of AD stages. 5 out of 9-miRNAs (hsa-miR-26a-5p, hsa-miR-181c-3p, hsa-miR-126-5p, hsa-miR-22-3p, hsa-miR-148b-5p) exhibited downregulation, and 1 (hsa-miR-106b-3p) showed upregulation in the mild stage of AD. Differential expression patterns of these 6 factors may be responsible for the initiation of the early stages of AD, even during the pre-symptomatic period. Individuals who are at early stages of AD but with higher risk of disease development could be determined by these biomarkers. For example, hsa-miR-22-3p exhibited higher sensitivity and specificity compared to other 5 miRNAs; hsa-miR-22-3p is highly expressed in the serum, thereby it has potential as a biomarker for early detection of AD. On the other hand, hsa-miR-6119-5p and hsa-miR-1246 were upregulated in the moderate, and hsa-miR-660-5p in the severe AD stage groups. Although it remains preliminary and requires future investigation, these three factors show clinical relevance as AD biomarkers in the later stages of the disease. But what is regrettable is that only a few miRNAs exhibited difference between AD subgroups. It is insufficient to monitor the progression of AD. All in all, the research on the circulating miRNAs as AD progression biomarkers continues and further compartmentalization of AD stages will confer benefits in such research.

Detecting miRNA expression levels as AD diagnostic tools gain considerable attention from researchers and clinicians; nevertheless, we do not advocate the sole use of miRNAs in the clinical settings for AD detection. Further investigations are required for validation of these factors in larger patient cohorts of patients with extensive statistical analyses. Combination of these factors with already established diagnosis methods will confer important advantages in obtaining more efficient and accurate results. Another possible benefit would be the evaluation of disease treatment by these factors. The mechanism behind the patho-histo-physiology of AD remains poorly understood [29]. The factors responsible for the disease initiation are not known. Parallel to such disease complexity, tools for AD diagnosis and classification of clinical stages still remain limited. In this context, we believe our findings provide an important approach and considerable potential in the diagnosis of AD.

Footnotes

ACKNOWLEDGMENTS

This work was supported by grant from the National Natural Science Foundation of China (31370897). The authors thank the participants in the study for making this work possible.

Authors’ disclosures available online ().

The supplementary material is available in the electronic version of this article: .

References

Ballard

, Gauthier

, Corbett

, Brayne

, Aarsland

, Jones

(2011) Alzheimer’s disease. Lancet 377, 1019–1031.

Mintun

, Larossa

, Sheline

, Dence

, Lee

, Mach

, Klunk

, Mathis

, DeKosky

, Morris

(2006) [11C]PIB in a nondemented population: Potential antecedent marker of Alzheimer disease. Neurology 67, 446–452.

Villemagne

, Burnham

, Bourgeat

, Brown

, Ellis

, Salvado

, Szoeke

, Macaulay

, Martins

, Maruff

, Ames

, Rowe

, Masters

, Australian Imaging

, Lifestyle Research

(2013) Amyloid beta deposition, neurodegeneration, and cognitive decline in sporadic Alzheimer’s disease: A prospective cohort study. Lancet Neurol 12, 357–367.

Schneider

, Hampel

, Buerger

(2009) Biological marker candidates of Alzheimer’s disease in blood, plasma, and serum. CNS Neurosci Ther 15, 358–374.

Kosik

(2006) The neuronal microRNA system. Nat Rev Neurosci 7, 911–920.

Djuranovic

, Nahvi

, Green

(2011) A parsimonious model for gene regulation by miRNAs. Science 331, 550–553.

Bartel

(2004) MicroRNAs: Genomics, biogenesis, mechanism, and function. Cell 116, 281–297.

Hausler

, Keller

, Chandran

, Ziegler

, Zipp

, Heuer

, Krockenberger

, Engel

, Honig

, Scheffler

, Dietl

, Wischhusen

(2010) Whole blood-derived miRNA profiles as potential new tools for ovarian cancer screening. Br J Cancer 103, 693–700.

Liu

, Zhang

, Hu

, Li

, Wang

, Yang

, Huang

, Chen

, Zhang

, Zhuang

, Deng

, Liu

, Yin

, Wang

, Zen

, Ba

, Zhang

(2011) A five-microRNA signature identified from genome-wide serum microRNA expression profiling serves as a fingerprint for gastric cancer diagnosis. Eur J Cancer 47, 784–791.

10.

, Chong

, Jin

, Lam

, Shin

, Yu

, Poon

, Ng

, Sung

(2009) Differential expression of microRNAs in plasma of patients with colorectal cancer: A potential marker for colorectal cancer screening. Gut 58, 1375–1381.

11.

Sayed

, Hong

, Chen

, Lypowy

, Abdellatif

(2007) MicroRNAs play an essential role in the development of cardiac hypertrophy. Circ Res 100, 416–424.

12.

Fichtlscherer

, De Rosa

, Fox

, Schwietz

, Fischer

, Liebetrau

, Weber

, Hamm

, Roxe

, Muller-Ardogan

, Bonauer

, Zeiher

, Dimmeler

(2010) Circulating microRNAs in patients with coronary artery disease. Circ Res 107, 677–684.

13.

, Hannon

(2004) MicroRNAs: Small RNAs with a big role in gene regulation. Nat Rev Genet 5, 522–531.

14.

Fineberg

, Kosik

, Davidson

(2009) MicroRNAs potentiate neural development. Neuron 64, 303–309.

15.

Lau

, Bossers

, Janky

, Salta

, Frigerio

, Barbash

, Rothman

, Sierksma

, Thathiah

, Greenberg

, Papadopoulou

, Achsel

, Ayoubi

, Soreq

, Verhaagen

, Swaab

, Aerts

, De Strooper

(2013) Alteration of the microRNA network during the progression of Alzheimer’s disease. EMBO Mol Med 5, 1613–1634.

16.

Valadi

, Ekstrom

, Bossios

, Sjostrand

, Lee

, Lotvall

(2007) Exosome-mediated transfer of mRNAs and microRNAs is a novel mechanism of genetic exchange between cells. Nat Cell Biol 9, 654–659.

17.

Hunter

, Ismail

, Zhang

, Aguda

, Lee

, Yu

, Xiao

, Schafer

, Lee

, Schmittgen

, Nana-Sinkam

, Jarjoura

, Marsh

(2008) Detection of microRNA expression in human peripheral blood microvesicles. PLoS One 3, e3694.

18.

Cheng

, Doecke

, Sharples

, Villemagne

, Fowler

, Rembach

, Martins

, Rowe

, Macaulay

, Masters

, Hill

, Australian Imaging, Biomarkers, Lifestyle (AIBL) Research Group (2015) Prognostic serum miRNA biomarkers associated with Alzheimer’s disease shows concordance with neuropsychological and neuroimaging assessment. Mol Psychiatry 20, 1188–1196.

19.

Galimberti

, Villa

, Fenoglio

, Serpente

, Ghezzi

, Cioffi

, Arighi

, Fumagalli

, Scarpini

(2014) Circulating miRNAs as potential biomarkers in Alzheimer’s disease. J Alzheimers Dis 42, 1261–1267.

20.

Kumar

, Dezso

, MacKenzie

, Oestreicher

, Agoulnik

, Byrne

, Bernier

, Yanagimachi

, Aoshima

, Oda

(2013) Circulating miRNA biomarkers for Alzheimer’s disease. PLoS One 8, e69807.

21.

Burgos

, Malenica

, Metpally

, Courtright

, Rakela

, Beach

, Shill

, Adler

, Sabbagh

, Villa

, Tembe

, Craig

, Van Keuren-Jensen

(2014) Profiles of extracellular miRNA in cerebrospinal fluid and serum from patients with Alzheimer’s and Parkinson’s diseases correlate with disease status and features of pathology. PLoS One 9, e94839.

22.

McKhann

, Drachman

, Folstein

, Katzman

, Price

, Stadlan

(1984) Clinical diagnosis of Alzheimer’s disease: Report of the NINCDS-ADRDA Work Group under the auspices of Department of Health and Human Services Task Force on Alzheimer’s Disease. Neurology 34, 939–944.

23.

Leidinger

, Backes

, Deutscher

, Schmitt

, Mueller

, Frese

, Haas

, Ruprecht

, Paul

, Stahler

, Lang

, Meder

, Bartfai

, Meese

, Keller

(2013) A blood based 12-miRNA signature of Alzheimer disease patients. Genome Biol 14, R78.

24.

Blondal

, Jensby Nielsen

, Baker

, Andreasen

, Mouritzen

, Wrang Teilum

, Dahlsveen

(2013) Assessing sample and miRNA profile quality in serum and plasma or other biofluids. Methods 59, S1–S6.

25.

Kirschner

, Edelman

, Kao

, Vallely

, van Zandwijk

, Reid

(2013) The Impact of Hemolysis on Cell-Free microRNA Biomarkers. Front Genet 4, 94.

26.

Singh

, Preus

, Hu

, Yan

, Long

, Morrison

, Nesline

, Johnson

, Koochekpour

, Kohli

, Liu

, Trump

, Sucheston-Campbell

, Campbell

(2014) Serum microRNA expression patterns that predict early treatment failure in prostate cancer patients. Oncotarget 5, 824–840.

27.

Koberle

, Kronenberger

, Pleli

, Trojan

, Imelmann

, Peveling-Oberhag

, Welker

, Elhendawy

, Zeuzem

, Piiper

, Waidmann

(2013) Serum microRNA-1 and microRNA-122 are prognostic markers in patients with hepatocellular carcinoma. Eur J Cancer 49, 3442–3449.

28.

Gillis

, Rijlaarsdam

, Eini

, Dorssers

, Biermann

, Murray

, Nicholson

, Coleman

, Dieckmann

, Belge

, Bullerdiek

, Xu

, Bernard

, Looijenga

(2013) Targeted serum miRNA (TSmiR) test for diagnosis and follow-up of (testicular) germ cell cancer patients: A proof of principle. Mol Oncol 7, 1083–1092.

29.

Sheinerman

, Umansky

(2013) Circulating cell-free microRNA as biomarkers for screening, diagnosis and monitoring of neurodegenerative diseases and other neurologic pathologies. Front Cell Neurosci 7, 150.