A combined miRNA-piRNA signature to detect Alzheimer's disease

Abstract

Alzheimer's disease (AD) is the most common neurodegenerative disorder causing huge emotional and economic burden to our societies. An effective therapy has not been implicated yet, which is in part also due to the fact that pathological changes occur years before clinical symptoms manifest. Thus, there is a great need for the development of a translatable biomarker. Recent evidence highlights microRNAs as candidate biomarkers. In this study, we use next-generation sequencing to study the small noncoding RNAome (sncRNAome) in exosomes derived from human cerebrospinal fluid (CSF). We show that the sncRNAome from CSF-derived exosomes is dominated not only by microRNAs (miRNAs) but also by PIWI-interacting RNAs (piRNAs). We define a combined signature consisting of three miRNAs and three piRNAs that are suitable to detect AD with an AUC of 0.83 in a replication cohort and furthermore predict the conversion of mild-cognitive impaired (MCI) patients to AD dementia with an AUC of 0.86 for the piRNA signature. When combining the smallRNA signature with pTau and Aβ 42/40 ratio the AUC reaches 0.98. Our study reports a novel exosomal small noncoding RNA signature to detect AD pathology and provides the first evidence that in addition to miRNAs, piRNAs should also be considered as a candidate biomarker for AD.

Fig. 1. Analysis of the exosomal sncRNAome.

a Exosomes isolated from human CSF were analyzed via EM (upper left panel), for fragment size by using a nanosight instrument (right panel) and via immunoblot for exosomal marker proteins (lower panel). b Electropherogram showing the profile of RNA isolated from exosomes. c Electropherogram showing the profile of RNA isolated from exosome-free CSF. d Electropherogram showing the profile of RNA isolated from lysed exosomes treated with DNAase (left) and RNAase (right). e Left panel: Pie chart showing the distribution of small noncoding RNAs in human CSF exosomes. Pie chart on the top right shows the genomic distribution of the human piRNAome for comparison. The lower right pie chart shows genomic annotation of the human CSF exosomal piRNAome. Note that in contrast to the entire piRNAome (upper right pie chart), the majority of piRNAs reside in the first exon of coding genes. f Top 5 expressed miRNAs (blue) and piRNAs (red) in human CSF exosomes. g Heatmap showing expression values of the 3-p and 5-p arms of all miRNAs detected in human CSF and in the human cortex (Brodmann Area 9). h Graphs showing Pearson correlation between miRNA (two left panels) and piRNA (two right panels) expression values of hippocampal and cortical neurons vs. the corresponding miRNA and piRNA expression in exosomes released from these cells

Fig. 2. A sncRNA signature to diagnose AD patients.

a Demographic information of the human cohorts used for signature identification and testing. b Measure of Relevance (MoR) analysis for miRNA differences between AD and control samples in the signature cohort. The dotted red line represents the critical MoR value cut off. miRNAs above the dotted red line are considered informative. miRNAs marked in blue are not confounded by age and gender after MANCOVA analysis. The inset shows fold-change (log2) value of the three identified miRNAs between control and AD patients. Their significance level (Bonferroni-corrected p value < = 0.05) with their significance codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘’ 1 is shown on the top of the bars. c Measure of Relevance (MoR) analysis for piRNA differences between AD and control samples in the signature cohort. The dotted red line represents the critical MoR value cutoff. piRNAs above the dotted red line are considered informative. piRNAs marked in red are not confounded by age and gender after MANCOVA analysis. The inset shows the fold-change (log2) value of the three identified piRNAs between control and AD patients. Their significance level (Bonferroni-corrected p value ≤0.05) with their significance codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 is shown on the top of the bars. d Receiver-operating characteristic (ROC) plot was obtained during the performance testing by using pTau levels and Aβ42/40 ratio obtained from CSF samples of the signature cohort on the replication cohort. Training was performed on the signature cohort with a tenfold cross-validation. The inset plot shows the variable importance. e Heatmap showing Pearson correlation coefficient between normalized expression of the sncRNA signature, pTau, and Aβ42/40 ratio in the signature cohort. Note that the sncRNA signature does not correlate significantly with pTau levels and Aβ42/40 ratio. f ROC showing performance of the six sncRNA signatures when tested on the replication cohort. Training was done on the signature cohort with a tenfold cross-validation. The inset plot shows the variable importance of the six individual sncRNAs. g ROC showing performance of the combined 6 sncRNA signatures with pTau and Aβ42/40 ratio levels on the replication. A mean AUC of 0.98 was obtained. F, female; M, male

Fig. 3. A sncRNA signature to predict conversion of MCI patients.

a Demographic information of the DCN longitudinal and multicenter cohort of MCI. In total, baseline CSF exosomes from 17 participants diagnosed with MCI were analyzed. At the 10-year follow-up six individuals had converted to dementia (3 male and 3 female). b ROC shows performance of our three piRNA signatures (identified in the signature cohort) on the DCN cohort consisting of converting and stable MCI participants. c ROC with the mean AUC of 0.96 was obtained by using a combination of our piRNA signature with pTau and Aβ42/40 ratio values. Error bars indicate SD

Fig. 4. Performance of the CSF sncRNA signature in classifying patients on the basis of postmortem brain tissue.

a Demographic information for postmortem brain tissue samples included in the analysis (published dataset GSE48552). b The six sncRNA signatures defined via the analysis of the signature cohort were tested on the data obtained from postmortem brain tissue (published brain cohort). ROC reveals a mean AUC of 0.89 suggesting that the sncRNA signature obtained from CSF helps to diagnose AD patients based on sncRNA expression in postmortem brain tissue. c Upper panel shows the confirmed target genes of the three miRNAs that are part of our sncRNA signature. Lower panel shows the significantly enriched signaling pathways based on the confirmed targets of the three miRNAs that are a part of the sncRNA signature. Error bars indicate SD