We have organized the results of this study into four sections: i) the selection of studies from the systematic review, ii) the individual exploratory analysis carried out on each study, iii) the differential miRNA expression profiles from several comparisons, and iv) integration of differential miRNA expression results with a meta-analysis approach and the functional enrichment developed within a Gene Set Enrichment Analysis (GSEA) methodology (Fig. 1).
Systematic Review
We developed a systematic review following PRISMA guidelines to identify suitable miRNA-focused expression studies in human AD. Studies must include information regarding the sex associated with each sample. Following the search criteria (Fig. 2), we initially identified twenty-seven studies (twenty-six in GEO and one in ArrayExpress). Following the removal of duplicate studies (n = 2), non-human studies (n = 4), non-AD-focused studies (n = 9), studies with non-suitable experimental designs (n = 2), and non-miRNA-based studies (3), we selected seven studies as initially suitable for further analysis. Out of those seven studies, we left GSE63501 and GSE153284 out of the analysis due to the impossibility of access to the expression data and a lack of standardization, respectively. Therefore, we included five studies in the analysis: GSE157239 [47], GSE16759 [48], and GSE48552 [49] for brain tissue and GSE120584 [50] and GSE46579 [51] for blood samples (Table 1).
Table 1
Summary of the selected studies from the systematic review
ID | Study type | Platform | Samples | Control | Cases | Type of sample | PMID |
GSE157239 | Non-coding RNA profiling by array | GPL21572 | 16 | 8 | 8 | Brain (Temporal cortex) | 32920076 |
GSE16759 | Non-coding RNA profiling by array | GPL8757 | 8 | 4 | 4 | Brain (Parietal lobe) | 20126538 |
GSE48552 | Non-coding RNA profiling by high-throughput sequencing | GPL11154 | 12 | 6 | 6 | Brain (Prefrontal cortex) | 24014289 |
GSE120584 | Non-coding RNA profiling by array | GPL21263 | 1309 | 288 | 1021 | Blood (serum) | 34686734 |
GSE46579 | Non-coding RNA profiling by high-throughput sequencing | GPL11154 | 70 | 22 | 48 | Blood (whole blood) | 23895045 |
Exploratory and Differential Expression Analysis
The exploratory analysis allowed an assessment of the distribution of the expression patterns in each study and to update the annotation of the miRNAs analyzed. We identified common miRNAs between studies and only considered those appearing in two or more studies for the integration analysis. Condition and sex distribution of the samples (Fig. 3) skewed towards female patients in most studies, with a lower degree of male representation in control or AD groups. GSE120584 possessed a much higher sample size than the other selected studies (n = 1309).
Expression data distribution did not provide evidence of anomalous samples, the hierarchical clustering of the samples did not display absolute divisions of the samples based on any of the experimental conditions, and the PCA visualization provided no evidence of bias.
The differential expression analysis of the individual studies returned profiles describing the altered expression of miRNAs in female and male patients in GSE120584, GSE46579, and GSE48552 studies but not in GSE16759 and GSE157239 (Table 2). Subsequent analyses focused on integrating individual differential expression analysis to obtain robust results regarding the alteration of miRNAs for the comparisons considered.
Table 2
Differentially expressed miRNAs by study and by comparison
Tissue | Study | miRNAs Analyzed | Female Comparison | Male Comparison |
Upregulated | Downregulated | Upregulated | Downregulated |
Brain | GSE157239 | 2561 | 0 | 0 | 0 | 0 |
GSE16759 | 462 | 0 | 0 | 0 | 0 |
GSE48552 | 832 | 88 | 79 | 34 | 7 |
Blood | GSE120584 | 2521 | 404 | 47 | 491 | 553 |
GSE46579 | 326 | 37 | 58 | 7 | 16 |
Meta-analysis and Functional Enrichment
We performed four meta-analyses, integrating the differential miRNA expression results of sets of studies based on sample tissue (blood or brain) and sex (females and males) (Table 3). We performed a random-effects DerSimonian-Laird (DL) meta-analysis on each combination's differential miRNA expression results from the limma/limma + voom approach. We obtained a combined logFC for each miRNA under analysis and an associated BH-adjusted p-value.
Table 3
Results of differential expression meta-analyses based on individual studies
Tissue | miRNAs Analyzed | Female Comparison | Male Comparison |
Upregulated | Downregulated | Upregulated | Downregulated |
Brain | 850 | 1 | 5 | 2 | 2 |
Blood | 314 | 16 | 0 | 18 | 4 |
Blood meta-analyses
We found significantly altered miRNAs in female and male AD patients via meta-analyses based on blood samples (Fig. 4). Sixteen miRNAs became significantly overexpressed in female AD patients compared to female control patients (Fig. 4A), while eighteen miRNAs became significantly overexpressed and four underexpressed in male AD patients compared to male control patients (Fig. 4B). We intersected profiles of overexpressed miRNAs in female and male AD patients for comparative purposes, which revealed a common increase in the expression of nine miRNAs in AD patients of both sexes and the exclusive overexpression of seven miRNAs in females and nine miRNAs in males (Fig. 4C). Then, we compared the expression profiles of these altered miRNAs in male and female AD patients. miRNAs altered in female AD patients shared overall similar expression patterns in females and males (Fig. 4D). miRNAs altered exclusively in male AD patients mainly shared similar expression patterns in males and females (Fig. 4E); however, hsa-mir-145-5p displayed a significant decrease in males but did not change in females.
We next compared the target profiles of the differentially expressed miRNAs to unveil those genes that may be impacted by AD (Supplementary Table 1). The top genes targeted by miRNAs significantly altered in female AD patients included ANKRD52 (target of eight miRNAs), CELF1 and LARP1 (target of seven miRNAs), CBX6, KMT2D, SETD5, SRCAP, SRRM2, and TAOK1 (target of six miRNAs). The top genes targeted by miRNAs significantly altered in male AD patients included LARP1, FUS, BAZ2A, KMT2D, and DICER1; the three miRNAs underexpressed in male AD patients targeted BTBD3, NDN, NUP43, PIK3C2B, RAC1, RASA1, RCAN2, RNF38, and RPRM.
Based on the complete miRNA transcriptomic profile, we performed a GSEA on the biological process (BP) ontology of GO terms (Fig. 5) using gene targets of all miRNAs with altered expression in AD patients regardless of their directionality. This selection allows us to identify BPs affected by the upregulation and/or downregulation of distinct genes. The functional enrichment analysis in AD females revealed six altered BPs (Fig. 5A); the 'protein polyubiquitination' and 'response to transforming growth factor beta' terms became downregulated while the 'positive regulation of cytoplasmic translation,' 'cytoplasmic translation,' 'mRNA splicing, via spliceosome' and 'RNA splicing' terms became upregulated. Meanwhile, we found 351 affected BP terms in male AD patients (Supplementary Table 2); nine BP terms mainly related to sensory perception and ion transmembrane transport increased, while 342 BP terms decreased. Supplementary Table 3 and Fig. 5C summarize the BPs affected and the top ten clusters and their parent terms, respectively.
Brain meta-analyses
The brain meta-analyses revealed the altered expression of miRNAs in female and male AD patients (Fig. 6). Female AD patients displayed the underexpression of five miRNAs and the overexpression of one miRNA (Fig. 6A); meanwhile, male AD patients displayed the underexpression of two miRNAs and the overexpression of two miRNAs (Fig. 6B). The intersection of the altered miRNAs in female and male AD patients revealed five exclusively altered in females, three exclusively altered in males, and one common underexpressed miRNA (hsa-miR-767-5p). Those miRNAs altered exclusively in female AD patients shared similar expression patterns in both sexes except for hsa-miR-105-3p, which displayed an increase in female and a slight decrease in male AD patients (Fig. 6C). Those miRNAs altered exclusively in male AD patients also shared similar expression patterns in females except for hsa-mir-3149, which displayed an increase in male and a non-significant decrease in female AD patients (Fig. 6D).
We then explored the target genes of the differentially expressed miRNAs in female and male AD patients (Supplementary Table 4). The genes targeted by the miRNA significantly increased in female AD patients (hsa-miR-105-3p) included CBLN2, GOLIM4, and UHMK1 and the two miRNAs significantly decreased (hsa-miR-431-3p and hsa-miR-767-5p) included MDM2, MTCH2, and MTRNR2L1. The genes targeted by the two miRNAs significantly increased in male AD patients (hsa-miR-491-3p and hsa-miR-3149) included NOM1 and ZNF226, while the genes targeted by the two miRNAs significantly decreased represented a list of seventy-one genes. Notably, the functional enrichment performed on the miRNA profiles of female and male AD patients failed to reveal any specifically affected BP terms.