MicroRNA Expression Levels Are Altered in the Cerebrospinal Fluid of Patients with Young-Onset Alzheimer’s Disease
- 1.2k Downloads
Clinical diagnosis of Alzheimer’s disease (AD) prior to the age of 65 years is classified as young-onset (YOAD), whereas diagnosis after the age of 65 years is considered late-onset (LOAD). Although rare autosomal mutations more commonly associate with YOAD, most YOAD and LOAD cases are sporadic. YOAD and LOAD share amyloid and tau pathology, but many YOAD patients show increased disease severity and rate of progression. The current study examined the microRNA (miRNA) expression profile from exosomes isolated from the cerebrospinal fluid (CSF) of YOAD patients with biomarker-confirmed AD. Results uncovered miR-16-5p, miR-125b-5p, miR-451a, and miR-605-5p as differentially expressed in the CSF-derived exosomes of YOAD patients when compared with healthy controls (HC). In a cohort of LOAD patients, miR-125b-5p, miR-451a, and miR-605-5p were similarly altered in expression, but miR-16-5p showed similar expression to control. Analysis of the mRNA targets of these miRNAs revealed transcripts enriched in biological processes relevant to the post-mortem posterior cingulate cortex transcriptome in YOAD from a previously published microarray study, including those related to neuron projections, synaptic signaling, metabolism, apoptosis, and the immune system. Hence, these miRNAs represent novel targets for uncovering disease mechanisms and for biomarker development in both YOAD and LOAD.
KeywordsAlzheimer’s disease Young-onset Cerebrospinal fluid Exosomes MicroRNA
Young-onset Alzheimer’s disease (YOAD) occurs before the age of 65 years and represents up to 10% of all AD cases. Fewer than 1% of AD are caused by autosomal dominant mutations [1, 2, 3]; hence, the majority of AD cases are sporadic regardless of age at disease onset [4, 5]. Although an age of 65 years for delineating YOAD versus late-onset AD (LOAD) was selected based on social factors such as the traditional retirement age , key imaging, neuropathological, and neurochemical differences based on an age cutoff of 65 years have been reported in the literature. YOAD patients present with more severe gray matter atrophy , more abundant senile plaques, neurofibrillary tangles, and synaptic loss , as well as greater deficits in acetylcholine  than LOAD patients. YOAD patients appear more likely to present with impaired language, attention, and visuospatial function, compared with LOAD patients who experience more memory deficits .
In terms of rate of progression, some reports have shown a faster rate of cognitive decline in younger patients [8, 10, 11, 12]; others found no association between the age of onset and rate of decline [13, 14, 15]. Currently, biomarker protein levels of amyloid β-42 (Aβ42), total tau (t-tau), and phosphorylated tau (p-tau) in the cerebrospinal fluid (CSF) of YOAD and LOAD patients are routinely used to confirm a diagnosis consistent with AD [16, 17, 18, 19]. Some studies have reported a higher level of CSF t-tau and p-tau in YOAD compared with LOAD, suggesting a more severe disease and more rapid decline [20, 21]. However, this difference in CSF tau levels is inconsistently found . Other novel biomarkers that relate to disease mechanism are of great interest in YOAD as current biomarkers of amyloid and tau do not provide differentiating features between YOAD and LOAD.
MiRNAs are stable, small, non-coding RNAs (21–23 nucleotides) involved in the degradation and/or translational repression of target messenger RNAs (mRNAs) (reviewed in ). Up to 2588 mature human miRNAs (miRNAs) have been identified (miRbase.org). Individual miRNAs can target and silence up to thousands of mRNAs, and multiple miRNAs can target single genes (reviewed in ). Cell-to-cell communication is considered a key role for miRNAs since they are released into the extracellular milieu by several mechanisms, including being complexed with Argonaute [25, 26] and lipoprotein particles  and packed into small extracellular vesicles called exosomes for exchange of genetic material between cells [26, 28]. Exosomes in particular represent a stable and enriched source of miRNAs in biofluids such as the blood [29, 30] and CSF . Importantly, exosomes have been implicated in cell-to-cell communication within the central nervous system. Exosomes secreted from neurons can regulate the brain vasculature , and exosomes secreted from astrocytes can modulate synaptic plasticity [33, 34]. In AD, neuronal exosomes may be involved in Aβ42 release as a result of early endosomal maturation  but may also be involved in intracerebral uptake of Aβ42 . Exosomes derived from AD and Down syndrome blood contain lower Aβ42 and increased p-tau  reflecting the changes seen in CSF of AD . Hence, we reasoned that specifically exploring miRNA expression profiles in exosomes would inform on AD-relevant disease mechanisms.
Generally, studies elucidating the miRNA expression profile in the CSF of AD patients are currently surging in the literature . In the majority of these studies, LOAD was compared with healthy controls [39, 40, 41, 42, 43]; others compared LOAD with healthy controls and other neurological diseases [44, 45, 46], while LOAD was also examined against other forms of dementia [47, 48]. Whole CSF was examined in most cases, whereas two of these studies profiled for miRNA changes in CSF-derived exosomes from LOAD patients. The first study used TaqMan miRNA arrays (746 human miRNAs) to profile for changes in the CSF from Parkinson’s disease (PD) and LOAD patients relative to controls . While several candidate miRNAs were uncovered for both PD and LOAD, validation by independent real-time PCR was only conducted on the PD cohort. In the second study , LOAD was compared with healthy controls in whole CSF using Exiqon’s human miRNome panels (752 human miRNAs) followed by examination of candidates in CSF-derived exosomes. Several candidate miRNAs were altered in LOAD compared with controls. In CSF-derived exosomes, the detectability of miRNAs increased, especially when testing the same candidates that were already identified in the whole CSF screen. However, the specific, high-throughput miRNA profile changes occurring in the CSF-derived exosomes of YOAD remain to be elucidated.
The current study explores the miRNA expression profile in exosomes derived from the CSF of biomarker-confirmed sporadic YOAD patients compared with that in healthy controls (HC). Here, we uncovered a decrease in miR-16-5p, miR-451a, and miR-605-5p and an increase in miR-125b-5p in YOAD patients versus HC. Combining the relative expression of these four miRNAs by regression analysis effectively distinguished YOAD relative to HC. Interestingly, the four miRNAs altered in YOAD share common targets and pathways altered in the post-mortem YOAD brain . In a cohort of LOAD patients, we showed that miR-451a and miR-605-5p were similarly decreased and miR-125b-5p increased in LOAD but there was no significant difference in miR-16-5p expression compared with HC. Hence, these results revealed three miRNAs with altered expression in the CSF-derived exosomes of both YOAD and LOAD. The YOAD-specific decrease of exosomal miR-16-5p provides a potential candidate involved in disease mechanisms related to YOAD.
Materials and Methods
Patient and Clinical Assessment
All patients with AD were seen at the University Health Network (UHN) Memory Clinic (Toronto Western Hospital) between 2011 and 2016 and diagnosed with possible or probable AD (McKhann, 2011). The cognitively normal healthy control CSF samples (n = 12) were obtained from UHN (n = 2) and University of California San Francisco Memory and Aging Center (n = 10). The YOAD group comprised patients aged less than 65 years (n = 17) and the LOAD group greater than 65 years (n = 13). The lumbar punctures were performed according to ADNI protocol , and CSF was collected in polypropylene tubes. A clinical AD diagnosis was confirmed using the CSF protein biomarkers: CSF Aβ42, p-tau, and t-tau levels. Innogenetic assays were run, and the results were considered consistent with AD diagnosis if p-tau > 68 pg/ml and Aβ42 to t-tau index (ATI) < 0.8 [18, 51, 52]. Patients also underwent cognitive assessment using either the Montreal Cognitive Assessment  or modified Behavioural Neurological Assessment .
Isolation of Exosomal miRNA from CSF
Exosomal preparations were performed using the miRCURY™ Exosome Isolation Kit (Exiqon) following the manufacturer’s instructions. To prepare each sample, 1.1 ml of undiluted input CSF was centrifuged for 5 min at 3000×g to pellet cell debris and 1.0 ml of the CSF supernatant was used as input for exosome extraction. High-quality miRNA was isolated from each exosome prep along with appropriate spike-in controls (Exiqon). Next, column purification was performed using the miRCURY RNA Isolation Kit following the manufacturer’s instructions (Exiqon).
Discovery Phase: High-Throughput PCR with Exiqon Human miRNome Panels I + II
Complementary DNA (cDNA) was synthesized using the locked nucleic acid (LNA) Universal cDNA Synthesis Kit (Exiqon). The ExiLENT SYBR Green 2X Master Mix (Exiqon) was used to prepare cDNA samples for amplification and visualization by quantitative real-time PCR (qrt-PCR). For each sample, cDNA was added to the SYBR master mix and was loaded at 10 μl per well across Exiqon human miRNome panels I + II (V4.M, Exiqon), which are 2 × 384-well plates consisting of a total of 752 well-established miRNA human primer sets. Both 384-well plates were run in tandem on a 7900HT thermocycler (Applied Biosystems, Life Technologies).
Quality Control, Normalization, and Statistical Analyses
For data filtering and quality control or individual reactions, raw amplification and melting curve data obtained for both Exiqon human panels I + II on the 7900HT thermocycler were imported into the Thermo Fisher Cloud Relative Quantification (RQ) app (Thermo Fisher Scientific, https://apps.thermofisher.com/apps/dashboard/#). Through automated processing and visual inspection across plates, only reaction wells displaying linear amplification, Ct values < 39, and that passed a melt curve analysis were included in subsequent analysis. Subsequently, all human panel data from both YOAD (n = 17 × 2 plates) and HC (n = 12 × 2 plates) was simultaneously imported into the GenEx software (6.0) for sample-to-sample (inter-plate) calibration. The selection of miRNAs for normalization was performed using established algorithms geNorm  and Normfinder . From this, the spike-ins UniSp6 and cel-miR-39-3p as well as the stably and highly expressed endogenous miR-204-5p were used to normalize across all Ct values using a combined geometric mean of all three Ct values . A previous study showed miR-204-5p to be the most abundant miRNA, with highly stable expression in whole CSF (see Tables 4 and 5 in ). An ANOVA followed by pairwise comparisons was performed after normalization. Additional candidate miRNAs for validation were uncovered using presence/absence of signal data mining followed by Fisher’s exact test for significance. Visualization of relative miRNA expression was performed using custom R scripts and heatmap.2 function in the R package gplots (v3.0.1) and GraphPad Prism v7.0c.
Validation Phase: Individual Primer Set qrt-PCR
In the same YOAD (n = 17) and HC (n = 12) samples as the discovery, individual qrt-PCR reactions were performed to validate differentially expressed miRNAs in YOAD that were identified in the discovery phase. Validated miRNAs in YOAD were examined in a cohort of LOAD patients (n = 13). For each miRNA primer set, three technical replicates per sample were included. Using the same cDNA and SYBR prep as discussed above, qrt-PCR reactions were performed using individual LNA primer sets (Exiqon) on 96-well plates using the ABI Step One Plus Real-Time PCR System (Applied Biosystems, Life Technologies). A list of all primer sets used in the current study is in Supplementary Table 1. Data normalization was performed similarly as above with the geometric mean of UniSp6 and miR-204-5p. Relative miRNA expression changes were calculated as relative expression to control using the Ct value from each qrt-PCR reaction by the 2−ΔΔCt method [58, 59]. Statistical significance between groups was performed using one-way ANOVA followed by Bonferroni correction. All bar graphs were drawn in GraphPad Prism v7.0c.
Receiver operating characteristics (ROC) analysis was performed to evaluate the capacity of each individual validated miRNA to distinguish either YOAD or LOAD from HC. ROC analysis is an established statistical approach for assessing the diagnostic potential of a continuous clinical variable [60, 61]. In a ROC analysis, for a biomarker of disease, the cutoff for correctly identifying patients is called sensitivity (true positive rate) and is plotted against the specificity (false positive rate) to produce a ROC curve. To assess combinatorial performance of validated miRNAs, a generalized linear model (GLM) was fitted with the relative expression data from combinations of validated miRNAs. The clinical diagnosis for YOAD or LOAD was given binary outcomes (0 or 1, respectively), where the predicted probability was modeled with a binomial distribution and logit function . GLM data was produced using basic R (v3.3.2) functions. ROC curve probabilities, area under the curve (AUC), and CI were calculated using the R package pROC . Finally, k-fold cross validation of each GLM was performed using adapted code from the R package DAAG . All AUC data was estimated using the trapezoidal method, and all 95% confidence intervals (CI) for the probabilities calculated at each cutoff were calculated in R (v3.4.0) using bootstrap sampling  with 1000 bootstrap replicates.
MicroRNA Target Prediction
For miRNAs confirmed in the validation phase, target prediction was performed using TargetScan v7.1 . TargetScan predicts miRNA gene targets by considering both canonical and non-canonical miRNA binding sites on target mRNA using experimentally backed datasets. The overlap between miRNA targets was visualized using the R package venn (v1.2), which requires the R package QCA . Next, the tissue location of enriched expression for targets of validated miRNAs was determined using FunRich v3.0 which combines multiple established databases including UniProt, Human Protein Atlas, Human Proteome Browser, Human Proteome Map, ProteomicsDB, and Human Proteinpedia to infer regional and cell-type enrichment of target mRNAs for a given list of miRNAs .
Analysis of Published Microarray Data
Published raw microarray data from a post-mortem study that contrasted the posterior cingulate cortex (PCC) of sporadic YOAD (n = 7) and HC (n = 7) was retrieved from GEO accession ID GSE39420 . Several R packages available through the Bioconductor framework  were used to analyze the microarray data. A detailed pipeline for conducting a differential expression analysis on Affymetrix microarray data in R  was modified and applied here. Briefly, the Affymetrix fluorescent intensity data was normalized using the R package affy (v1.52.0)  and a differential expression analysis was performed using the R package limma (v3.30.13)  to distinguish the PCC of sporadic YOAD versus HC. The false discovery rate (FDR) was set at q < 0.05. Pathway analysis of overlapping transcripts was performed using the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway data from Database for Annotation, Visualization, and Integrated Discovery (DAVID v6.8) .
Gene Set Enrichment Analysis and Enrichment Map Visualization
Enrichment of Gene Ontology (GO) biological process (BP), cellular compartment (CC), and molecular function (MF) terms  was determined using Gene Set Enrichment Analysis (GSEA; v2.1.0) . Homo sapiens GO BP, CC, and MF gene sets without inferred electronic annotation from the February 2018 release were retrieved from the online repository available at http://download.baderlab.org/EM_Genesets/ . A log2 fold change ranked list of differentially expressed sporadic YOAD (n = 7) versus HC (n = 7) was imported into GSEA, and a GO BP analysis was performed. Results from the GO BP GSEA were imported into the Cytoscape  (v3.3.0) plug-in entitled Enrichment Map  (v2.1.0) to visualize GO BP, MF, and CC term themes in the data as a network of nodes and edges. AutoAnnotate was used to create clusters around redundant GO terms between nodes using similarity coefficients . To select the most overrepresented terms to report on the enrichment map, WordCloud  was used. FDR was set to q < 0.001 for all analyses. Gene sets were pre-ranked using GSEA.
Cerebrospinal Fluid Donor Characteristics
Cerebrospinal fluid donor demographic data
Age at time of LP (years)a
66.5 ± 7.7
Age of onset (years)a
56.8 ± 4.9
Age at time of LP (years)a
60.9 ± 4.6
Disease duration (years)a
3.9 ± 2.3
Age at time of testing (years)a
61.36 ± 4.7
12.1 ± 6.7 (min 3, max 20)
Age at time of testing (years)a
59.83 ± 4.5
Total score (/329)a
136.7 ± 62.7 (min 55, max 255)
6.5 ± 2.3
Memory immediate recall (/30)a
8.9 ± 4.6
Delayed recall (/27)a
2.8 ± 4.4
Delayed recognition (20)a
15.3 ± 3.7
15.7 ± 10.7
Executive function (/123)a
39.1 ± 31.6
48.3 ± 17.8
356.0 ± 159.1
Total tau (pg/ml)a
744.5 ± 375.0
101.7 ± 37.9
3 3 (8); 3 4 (6); 4 4 (2); 2 3 (1)
Age at time of LP (years)a
75.5 ± 4.6
Disease duration (years)a
3.6 ± 2.7
431.3 ± 139.4
Total tau (pg/ml)a
721.6 ± 245.1
97.1 ± 19.7
3 3 (6); 3 4 (6); 2 3 (1)
For the LOAD patients (n = 5 females; n = 8 males), the mean age at LP was 75.5 ± 4.6 years. The protein biomarker levels for the LOAD group were also consistent with AD (Aβ-42 = 431.3 ± 139.4 pg/ml; total tau = 721.6 ± 245.1 pg/ml; phospho-tau = 97.1 ± 19.7 pg/ml; ATI = 0.37 ± 0.17). In this case, prevalence of the ApoE4 genotypes in LOAD patients was 53.85% with zero alleles, 46.15% with one allele, and no patients with two alleles. The LOAD group presented with more disease comorbidities than the YOAD group (Supplementary Table 2). Overall, an average of 1.4 (min 0, max 5) disease comorbidities was observed across all YOAD patients with seven patients showing no comorbidities (Supplementary Table 2). LOAD patients showed an average of 2.6 (min 0, max 6) disease comorbidities, and only one patient showed no comorbidities (Supplementary Table 2). These results are consistent with previous findings showing that LOAD patients are likely to present with more disease comorbidities than YOAD [81, 82].
Discovery Phase: the miRNA Expression Profile from CSF-Derived Exosomes Is Altered in YOAD
Validation Phase: Distinct miRNAs Distinguish YOAD or LOAD from Healthy Controls
Next, we asked whether the expression of the validated miRNAs differed based on the age at which the LP was performed. Pearson correlation analysis was performed for the relative expression of all four validated miRNAs versus age at LP. We found that there was a significant negative correlation for miR-125b-5p expression in HC with age (Supplementary Fig. 4a; R2 = 0.624, p = 0.0022), but this effect was insignificant in YOAD (Supplementary Fig. 4b; R2 = 0.0055, p = 0.777) and LOAD (Supplementary Fig. 4c; R2 = 0.624, p = 0.0022). Hence, this suggests that expression of miR-125b-5p decreases with age in HC but may remain at high expression levels in AD regardless of age. Although there was a potential trend toward a positive correlation of miR-451a in LOAD (Supplementary Fig. 4c; R2 = 0.304, p = 0.063) and miR-605-5p in HC (Supplementary Fig. 4c; R2 = 0.2599, p = 0.091) and LOAD (Supplementary Fig. 4c; R2 = 0.263, p = 0.088), no other miRNAs showed an age effect.
For LOAD, the same ROC analysis was performed as stated above and summarized in Fig. 4b. Results for miR-125b-5p showed modest improvement at distinguishing LOAD from HC (AUC = 0.785; CI = 0.537−0.950) than with YOAD. The miR-451a showed high discriminatory potential to distinguish LOAD from HC (AUC = 0.847; CI = 0.679−0.956), but was not as robust as in YOAD. In the case of miR-605-5p, performance at distinguishing LOAD from HC (AUC = 0.765; CI = 0.491−0.913) was similar as in YOAD. Therefore, employing these miRNAs as binary predictors for LOAD suggests a performance ranking of miR-451a > miR-125b-5p > miR-605-5p.
Linear Combinations of Validated miRNAs Show Synergistic Performance Distinguishing YOAD or LOAD from Healthy Controls
Combinatorial performance of the relative expression of validated miRNAs to distinguish AD from HC
miR-16-5p, miR-125-5p, miR-451a, miR-605-5p
HC versus YOAD
miR-16-5p and miR-451a
HC versus YOAD
miR-125-5p miR-451a, miR-605-5p
HC versus LOAD
Validated miRNA Share Overlapping Targets and Inferred Regional Distribution
The putative mRNA targets of the four validated miRNAs in YOAD were uncovered using the TargetScan v7.1 algorithm . From this, 1508 mRNA targets for miR-16-5p, 100 targets for miR-125b-5p, 28 mRNA targets for miR-451a, and 4028 mRNA targets for miR-605-5p were found (Supplementary Fig. 5a). Overlap is evident between three or fewer groups of the four miRNAs, but no targets are shared between combinations of all four miRNAs (Supplementary Fig. 5a). For LOAD, no targets overlap for miR-125b-5p, miR-451a, and miR-605-5p together, but pairs of each of these miRNAs do share targets (Supplementary Fig. 5b). To assess brain region and cell-type specific localization of these miRNAs, we employed FunRich v3.0 . From this, mRNA targets were depleted in peripheral blood cells and the choroid plexus, as expected. Importantly, markers of the cerebral cortex, hippocampus, cerebellum, or simply brain were enriched. Although no targets were significantly enriched in cerebrospinal fluid, the percentage of predicted targets overlapping with cerebrospinal fluid was higher for all predicted targets overlapping with the blood, peripheral blood cells, blood vessels, and choroid plexus (Supplementary Fig. 5c). These results suggest predominantly CNS localization of transcripts targeted by the validated miRNAs. Analysis of these miRNAs with miRPath (v3.0) showed that combinations of at least four miRNAs overlapped in pathways such as regulating pluripotency of stem cells, PI3K-Akt signaling pathway, AMPK signaling pathway, adrenergic signaling pathway, neurotrophin signaling pathway, MAPK signaling pathway, and Wnt signaling pathway (data not shown).
Validated miRNA Targets Converge on Common Pathways with Published Microarray Data from the Posterior Cingulate Cortex of Sporadic YOAD
The current study revealed four miRNAs with altered expression in exosomes derived from CSF of YOAD patients: miR-16-5p, miR-125b-5p miR-451a, and miR-605-5p. In a cohort of LOAD patients, differential expression of miR-125b-5p, miR-451a, and miR-605-5p was also observed. The fact that miR-16-5p was unchanged in LOAD suggests that altered exosomal expression of miR-16-5p may differentiate YOAD from LOAD. Moreover, all four miRNAs altered in YOAD share putative targets and pathways relevant to the post-mortem YOAD brain transcriptome. Intriguingly, these mRNAs were overrepresented in previously implicated pathways in AD, including long-term potentiation [83, 84, 85], MAPK signaling [86, 87, 88], Wnt signaling [89, 90], axon guidance [91, 92], and calcium signaling pathway [93, 94, 95, 96, 97]. Further, the upregulation of GO terms related to apoptosis and immune response also recapitulates known alterations in AD [98, 99, 100, 101]. An overrepresentation of AD-relevant pathways suggests that these miRNAs are potentially combinatorial candidates for understanding disease mechanisms underlying AD. Considering that three of the four validated miRNAs are downregulated, this implicates the absence of target repression  by these miRNAs in the enriched pathways. In contrast, the opposite miRNA/mRNA regulatory relationship may be true for upregulated miR-125b-5p.
The significant increase in miR-125b-5p we observed in both YOAD and LOAD patients is in agreement with previous findings in LOAD showing an increase in several CNS structures [103, 104, 105, 106] and in CSF [40, 43]. Other reports have shown a decrease of miR-125b-5p in LOAD serum [47, 107] and in CSF [47, 108]. To our knowledge, we are the first to report increased expression levels of miR-125b-5p in CSF-derived exosomes from YOAD and LOAD patients. There is some evidence that miR-125b-5p has a microglial localization and may play a pro-inflammatory role . Moreover, miR-125b-5p overexpression results in hyperphosphorylation of tau and neurotoxicity in primary neuron cultures . However, future studies are required to further examine these mechanisms in the context of AD.
A striking depletion of miR-451a in both YOAD and LOAD was noted in the current study. Previous reports have shown miR-451a to be decreased in the hippocampus , temporal cortex , and CSF  in AD. MiR-451a has a potential role in AD pathogenesis through attenuation of ADAM10 expression . It appears that miR-451a is unchanged in LOAD serum  and is enriched in neuron expression profiles , suggesting that miR-451a loss in AD is a CNS-specific molecular signature. Since miR-451a is also decreased in amyotrophic lateral sclerosis  and major depressive disorder  and plays a role in peripheral inflammation , future experiments will confirm whether miR-451a is a general marker of neurodegeneration and/or neuroinflammation.
A previous microarray study reported a modest increase of miR-605-5p in blood mononuclear cells in LOAD , but the authors state that miR-605-5p has low expression in the blood. In the whole CSF fraction, Riancho and colleagues (2017) demonstrated expression in 9/10 HC and 9/10 LOAD samples , suggesting that miR-605-5p could be specifically altered in CSF-derived exosomes. The organ-specific function of miR-605-5p is not well established in the literature. One group has carefully shown that miR-605-5p promotes the P53 stress response and inhibits apoptosis in colorectal carcinoma cells . Another report recently showed evidence that miR-605-5p acts as a tumor suppressor in melanoma by inhibiting INPP4B . Future studies will explore whether there a CNS-specific role for miR-605-5p exists.
A drawback to our experimental pipeline is the exclusion of a discovery panel on the whole CSF fraction to act as a direct comparison to the CSF-derived exosomes. Hence, future studies will be required to show whether the miRNAs we identified are specifically altered in exosomes alone or are also in other compartments. One previous study suggested that both the extracellular Argonaute2-bound miRNA profile and the exosomal fraction are important, but they found the AUC was higher when the exosomal microRNA was used versus the Argonaute2 microRNA and so concluded that exosomal microRNA may in some way more accurately reflect the pathophysiology of temporal lobe epilepsy and status epilepticus . Despite this limitation, our results partially compare to a recent study exploring the miRNA expression profile in the CSF-derived exosomes from LOAD patients . For example, the study of Riancho and colleagues showed that miR-598 was not detected in the whole CSF fraction of AD patients but was detectable in the majority of CSF-derived exosome preparations. We detected miR-598 in all HC and YOAD patients in the current study, with no difference in relative expression between HC and YOAD. Riancho and colleagues also found that miR-9-5p was more likely to be detected in LOAD; however, we only detected this miRNA in 1/12 HC and 4/17 YOAD samples. Although, we used the same Exiqon exosome extraction kits and these differences could be attributable to CSF input volume (300 μl in Riancho et al. versus 1 ml in our study). Indeed, recent stoichiometric comparison of three commercial precipitation exosome extraction techniques versus ultracentrifugation  and subsequent miRNA cleanup  demonstrated the relationship between CSF volume input and the exosome particle quantity in the yield. The polymer-based exosome isolation kits demonstrate high yield and overlapping particle size distribution regardless of input . However, the co-precipitation of contaminants in the form of membrane fragments, aggregated proteins, lipoprotein complexes, and/or ribonucleoprotein particles represents caveats to the use of synthetic polymer buffers [122, 123].
We observed a decrease in miR-16-5p in the exosomes from CSF in YOAD. It is known that miR-16 is also a member of the miR-15/107 gene family, targets AD-specific mRNAs, and has been explored as a potential therapeutic target in early AD . Functional studies have uncovered miR-16-5p-mediated inhibition of APP, BACE1, MAPT, and NICASTRIN transcripts [124, 125] indicating that AD-relevant transcripts are directly regulated by miR-16-5p. Our findings here agree with previous observations showing a decrease of miR-16-5p in younger LOAD patient CSF compared to controls  and in older LOAD patient serum  and miR-16-2 in the CSF-derived exosomes from younger LOAD patients . Previous findings also showed that miR-16-5p is decreased in LOAD hippocampus but not in CSF  and its decrease correlated with early Braak staging .
Interestingly, we did not observe a decrease in miR-16-5p in the exosomes of CSF from our LOAD cohort. This finding may be attributable to differences in the profile of CSF-derived exosomes versus whole CSF, serum, and/or brain tissue. Another possibility is that there may be age-dependent factors contributing to miRNA expression in CSF-derived exosomes. The average age of the LOAD cohort at the time of LP in the current study (75.5 ± 4.6 years) versus the previous CSF study (69.5 ± 7.3 years) differs . In addition, differences exist between the average age of our LOAD and HC cohort (66.5 ± 7.7 years). Future studies will consider the longitudinal change of miR-16-5p expression and other miRNAs in AD CSF, serum, and brain tissue. Also, replication of these results in independent and other neurodegenerative disease cohorts will confirm whether a decrease in CSF exosome levels of these miRNAs is specific for younger AD patients.
Overall, this study uncovered a differential expression profile for both previously identified and novel miRNAs altered in AD and extended these findings to the exosomal compartment in CSF from YOAD and LOAD patients. The expression level of all four miRNAs effectively discriminates YOAD from HC, suggesting the potential for combinatorial value for detection of YOAD versus HC. Since these miRNAs target transcripts and pathways relevant to molecular processes underlying AD and provide some further understanding into the pathophysiological differences between YOAD and LOAD, further functional characterization of these miRNAs may offer new therapeutic avenues for patients with AD.
The authors wish to extend their gratitude to the patients and families involved in this study. Without their selfless contribution, this research would not be possible. Moreover, the authors thank Dr. Shangxi Xiao, Dr. Philip McGoldrick, and Taehyung Kim for their discussion and input.
PMM holds an Alzheimer Society of Canada Doctoral Award. RS holds an ALS Canada Clinical Fellowship and has received funding through the University of Toronto Eliot Phillipson Clinician Scientist Program. This research was supported by the Toronto Western and General Hospital Research Foundation, NIH grants U54NS092089 and R01AG038791, and by the Tau Consortium.
Compliance with Ethical Standards
Conflict of Interest
The authors declare that they have no conflict of interest.
- 2.Rogaev EI, Sherrington R, Rogaeva EA, Levesque G, Ikeda M, Liang Y, Chi H, Lin C et al (1995) Familial Alzheimer’s disease in kindreds with missense mutations in a gene on chromosome 1 related to the Alzheimer’s disease type 3 gene. Nature 376:775–778. https://doi.org/10.1038/376775a0 PubMedCrossRefGoogle Scholar
- 8.Sakai M, Hanyu H, Kume K, Sato T, Hirao K, Kanetaka H, Abe S, Kanaya K et al (2013) Rate of progression of Alzheimer’s disease in younger versus older patients: a longitudinal single photon emission computed tomography study: effect of age on AD progression. Geriatr Gerontol Int 13:555–562. https://doi.org/10.1111/j.1447-0594.2012.00934.x PubMedCrossRefGoogle Scholar
- 16.McKhann GM, Knopman DS, Chertkow H et al (2011) The diagnosis of dementia due to Alzheimer’s disease: recommendations from the National Institute on Aging-Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimers Dement 7:263–269. https://doi.org/10.1016/j.jalz.2011.03.005 PubMedPubMedCentralCrossRefGoogle Scholar
- 17.Sperling RA, Aisen PS, Beckett LA, Bennett DA, Craft S, Fagan AM, Iwatsubo T, Jack CR Jr et al (2011) Toward defining the preclinical stages of Alzheimer’s disease: recommendations from the National Institute on Aging-Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimers Dement 7:280–292. https://doi.org/10.1016/j.jalz.2011.03.003 PubMedPubMedCentralCrossRefGoogle Scholar
- 22.Bouwman FH, Schoonenboom NSM, Verwey NA, van Elk EJ, Kok A, Blankenstein MA, Scheltens P, van der Flier WM (2009) CSF biomarker levels in early and late onset Alzheimer’s disease. Neurobiol Aging 30:1895–1901. https://doi.org/10.1016/j.neurobiolaging.2008.02.007 PubMedCrossRefGoogle Scholar
- 25.Arroyo JD, Chevillet JR, Kroh EM, Ruf IK, Pritchard CC, Gibson DF, Mitchell PS, Bennett CF et al (2011) Argonaute2 complexes carry a population of circulating microRNAs independent of vesicles in human plasma. Proc Natl Acad Sci 108:5003–5008. https://doi.org/10.1073/pnas.1019055108 PubMedPubMedCentralCrossRefGoogle Scholar
- 34.Xin H, Wang F, Li Y, Lu QE, Cheung WL, Zhang Y, Zhang ZG, Chopp M (2017) Secondary release of exosomes from astrocytes contributes to the increase in neural plasticity and improvement of functional recovery after stroke in rats treated with exosomes harvested from microRNA 133b-overexpressing multipotent mesenchymal stromal cells. Cell Transplant 26:243–257. https://doi.org/10.3727/096368916X693031 PubMedPubMedCentralCrossRefGoogle Scholar
- 41.Riancho J, Vázquez-Higuera JL, Pozueta A, Lage C, Kazimierczak M, Bravo M, Calero M, Gonalezález A et al (2017) MicroRNA profile in patients with Alzheimer’s disease: analysis of miR-9-5p and miR-598 in raw and exosome enriched cerebrospinal fluid samples. J Alzheimers Dis 57:483–491. https://doi.org/10.3233/JAD-161179 PubMedCrossRefGoogle Scholar
- 42.Müller M, Kuiperij HB, Claassen JA, Küsters B, Verbeek MM (2014) MicroRNAs in Alzheimer’s disease: differential expression in hippocampus and cell-free cerebrospinal fluid. Neurobiol Aging 35:152–158. https://doi.org/10.1016/j.neurobiolaging.2013.07.005 PubMedCrossRefGoogle Scholar
- 44.Burgos K, Malenica I, Metpally R, Courtright A, Rakela B, Beach T, Shill H, Adler C et al (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. https://doi.org/10.1371/journal.pone.0094839 PubMedPubMedCentralCrossRefGoogle Scholar
- 48.Sørensen SS, Nygaard A-B, Christensen T (2016) miRNA expression profiles in cerebrospinal fluid and blood of patients with Alzheimer’s disease and other types of dementia—an exploratory study. Transl Neurodegener 5:6. https://doi.org/10.1186/s40035-016-0053-5 PubMedPubMedCentralCrossRefGoogle Scholar
- 49.Antonell A, Lladó A, Altirriba J, Botta-Orfila T, Balasa M, Fernández M, Ferrer I, Sánchez-Valle R et al (2013) A preliminary study of the whole-genome expression profile of sporadic and monogenic early-onset Alzheimer’s disease. Neurobiol Aging 34:1772–1778. https://doi.org/10.1016/j.neurobiolaging.2012.12.026 PubMedCrossRefGoogle Scholar
- 57.Wang W-X, Fardo DW, Jicha GA, Nelson PT (2017) A customized quantitative PCR microRNA panel provides a technically robust context for studying neurodegenerative disease biomarkers and indicates a high correlation between cerebrospinal fluid and choroid plexus microRNA expression. Mol Neurobiol 54:8191–8202. https://doi.org/10.1007/s12035-016-0316-2 PubMedCrossRefGoogle Scholar
- 64.Maindonald JH, Braun J (2010) Data analysis and graphics using R: an example-based approach, 3rd ed. Cambridge University Press, Cambridge; New YorkGoogle Scholar
- 67.Thiem A, Duşa A (2013) QCA: a package for qualitative comparative analysis. R J 5Google Scholar
- 70.Klaus B (2016) An end to end workflow for differential gene expression using Affymetrix microarrays. F1000Research 5Google Scholar
- 72.Smyth G (2005) limma: linear models for microarray data. Bioinforma Comput Biol Solut Using R Bioconductor:397–420Google Scholar
- 82.Haaksma ML, Vilela LR, Marengoni A, Calderón-Larrañaga A, Leoutsakos JMS, Olde Rikkert MGM, Melis RJF (2017) Comorbidity and progression of late onset Alzheimer’s disease: a systematic review. PLoS One 12:e0177044. https://doi.org/10.1371/journal.pone.0177044 PubMedPubMedCentralCrossRefGoogle Scholar
- 83.Shankar GM, Li S, Mehta TH, Garcia-Munoz A, Shepardson NE, Smith I, Brett FM, Farrell MA et al (2008) Amyloid-β protein dimers isolated directly from Alzheimer’s brains impair synaptic plasticity and memory. Nat Med 14:837–842. https://doi.org/10.1038/nm1782 PubMedPubMedCentralCrossRefGoogle Scholar
- 85.Palop JJ, Mucke L (2010) Amyloid-β-induced neuronal dysfunction in Alzheimer’s disease: from synapses toward neural networks. Nat Neurosci 13:812–818. https://doi.org/10.1038/nn.2583
- 90.Liu C-C, Tsai C-W, Deak F, Rogers J, Penuliar M, Sung YM, Maher JN, Fu Y et al (2014) Deficiency in LRP6-mediated Wnt signaling contributes to synaptic abnormalities and amyloid pathology in Alzheimer’s disease. Neuron 84:63–77. https://doi.org/10.1016/j.neuron.2014.08.048 PubMedPubMedCentralCrossRefGoogle Scholar
- 99.Gjoneska E, Pfenning AR, Mathys H, Quon G, Kundaje A, Tsai LH, Kellis M (2015) Conserved epigenomic signals in mice and humans reveal immune basis of Alzheimer’s disease. Nature 518:365–369. https://doi.org/10.1038/nature14252
- 100.Hardy J, Selkoe DJ (2002) The amyloid hypothesis of Alzheimer’s disease: progress and problems on the road to therapeutics. Science 297:353–356Google Scholar
- 101.Durrenberger PF, Fernando FS, Kashefi SN, Bonnert TP, Seilhean D, Nait-Oumesmar B, Schmitt A, Gebicke-Haerter PJ et al (2015) Common mechanisms in neurodegeneration and neuroinflammation: a BrainNet Europe gene expression microarray study. J Neural Transm 122:1055–1068. https://doi.org/10.1007/s00702-014-1293-0 PubMedCrossRefGoogle Scholar
- 103.Banzhaf-Strathmann J, Benito E, May S, Arzberger T, Tahirovic S, Kretzschmar H, Fischer A, Edbauer D (2014) MicroRNA-125b induces tau hyperphosphorylation and cognitive deficits in Alzheimer’s disease. EMBO J 33:1667–1680. https://doi.org/10.15252/embj.201387576 PubMedPubMedCentralCrossRefGoogle Scholar
- 110.Villela D, Ramalho RF, Silva ART, Brentani H, Suemoto CK, Pasqualucci CA, Grinberg LT, Krepischi ACV et al (2016) Differential DNA methylation of microRNA genes in temporal cortex from Alzheimer’s disease individuals. Neural Plast 2016:1–10. https://doi.org/10.1155/2016/2584940 CrossRefGoogle Scholar
- 111.Cheng C, Li W, Zhang Z, Yoshimura S, Hao Q, Zhang C, Wang Z (2013) MicroRNA-144 is regulated by activator protein-1 (AP-1) and decreases expression of Alzheimer disease-related a disintegrin and metalloprotease 10 (ADAM10). J Biol Chem 288:13748–13761. https://doi.org/10.1074/jbc.M112.381392 PubMedPubMedCentralCrossRefGoogle Scholar
- 113.Jovicic A, Roshan R, Moisoi N, Pradervand S, Moser R, Pillai B, Luthi-Carter R (2013) Comprehensive expression analyses of neural cell-type-specific miRNAs identify new determinants of the specification and maintenance of neuronal phenotypes. J Neurosci 33:5127–5137. https://doi.org/10.1523/JNEUROSCI.0600-12.2013 PubMedCrossRefGoogle Scholar
- 114.Maciotta S, Meregalli M, Torrente Y (2013) The involvement of microRNAs in neurodegenerative diseases. Front Cell Neurosci 7. https://doi.org/10.3389/fncel.2013.00265
- 116.Bandres E, Bitarte N, Arias F, Agorreta J, Fortes P, Agirre X, Zarate R, Diaz-Gonzalez JA et al (2009) microRNA-451 regulates macrophage migration inhibitory factor production and proliferation of gastrointestinal cancer cells. Clin Cancer Res 15:2281–2290. https://doi.org/10.1158/1078-0432.CCR-08-1818 PubMedCrossRefGoogle Scholar
- 117.Schipper HM, Maes OC, Chertkow HM, Wang E (2007) MicroRNA expression in Alzheimer blood mononuclear cells. Gene Regul Syst Biol 1:263Google Scholar
- 120.Raoof R, Jimenez-Mateos EM, Bauer S, Tackenberg B, Rosenow F, Lang J, Onugoren MD, Hamer H et al (2017) Cerebrospinal fluid microRNAs are potential biomarkers of temporal lobe epilepsy and status epilepticus. Sci Rep 7:3328. https://doi.org/10.1038/s41598-017-02969-6 PubMedPubMedCentralCrossRefGoogle Scholar
- 121.Helwa I, Cai J, Drewry MD, Zimmerman A, Dinkins MB, Khaled ML, Seremwe M, Dismuke WM et al (2017) A comparative study of serum exosome isolation using differential ultracentrifugation and three commercial reagents. PLoS One 12:e0170628. https://doi.org/10.1371/journal.pone.0170628 PubMedPubMedCentralCrossRefGoogle Scholar
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.