A spatio-temporal brain miRNA expression atlas identifies sex-independent age-related microglial driven miR-155-5p increase

Abstract

An in-depth understanding of the molecular processes composing aging is crucial to develop therapeutic approaches that decrease aging as a key risk factor for cognitive decline. Herein, we present a spatio-temporal brain atlas (15 different regions) of microRNA (miRNA) expression across the mouse lifespan (7 time points) and two aging interventions composed of 1009 samples. MiRNAs are promising therapeutic targets, as they silence genes by complementary base-pair binding of messenger RNAs and are known to mediate aging speed. We first established sex- and brain-region-specific miRNA expression patterns in young adult samples. Then we focused on sex-dependent and independent brain-region-specific miRNA expression changes during aging. The corpus callosum in males and the choroid plexus in females exhibited strong sex-specific age-related signatures. In this work, we identified three sex-independent brain aging miRNAs (miR-146a-5p, miR-155-5p and miR-5100). We showed for miR-155-5p that these expression changes are driven by aging microglia. MiR-155-5p targets mTOR signaling pathway components and other cellular communication pathways and is hence a promising therapeutic target.

Keywords: Aging; miRNA; transcriptional regulation.

Fig. 1:

Spatio-temporal overview and analysis of RNA expression in the aging mouse brain. a Study overview: Age stages of tissue collection at 3, 12, 15, 18, 21, 26 and 28 months in 15 different brain regions defined according to the Allen Brain atlas. The numbers below the timeline are indicating how many mouse individuals were observed per age stage and sex. Schematic plot created with BioRender.com. b The distribution of Spearman’s rank correlation coefficients between age and individual features for each RNA class. These correlations were calculated separately for each brain region using all available samples. The black solid line within each ridgeline indicates the median. On the right we see the number of features included in the ridgeline. c Boxplots showing the time series of tRNA tRNA-Glu-TTC-1-1 for three different brain regions each for male left and female right. The Spearman’s rank correlation coefficient from the miRNA expression with age is displayed above each plot. An asterisk indicates the significance level. d UMAP of all samples from all regions and for all tRNA features colored by regions as indicated in Fig. 1a. e Principal Variance Component Analysis for the tRNAs over all samples for the biological factors brain region, age, sex, and individual depicting the observed variance in percent. f UMAP of all samples from all regions and for all miRNA features colored by regions as indicated in Fig. 1a. g Principal Variance Component Analysis for the miRNAs again over all samples for the biological factors brain region, age, sex, and individual depicting the observed variance in percent. Colours are matching to the once from Fig. 1e.

Fig. 2:

Sex- and brain region-specific miRNA variation in the mouse brain. a Heatmaps of the 50 top miRNA from all brain regions determined by coefficient of variation calculated using the medians of the expression values of each brain region. Shown are the absolute standardized expression values (z-scores) for the younger male (left) and younger female (right) samples. The black borders are indicating the binarization (|z-score| > 0.5) on which a clustering into four clusters using a hierarchical clustering was applied. We consider a miRNA in a brain region different from the average brain if it surpasses the before mentioned threshold. If it only exceeds the threshold for one brain region, we call it brain region specific. The lines connecting the two heatmaps are highlighting the occurrence of common features (light green). If features are differing from the average brain for only one but the same brain region in both sexes, they are highlighted in pink and in dark green if they differ from the average brain in the same set of multiple brain regions. For visualization purposes, we removed features entirely below the selected threshold. b Venn diagram linking the features which are differing from the average brain per brain region between the male and the female heatmap. c Principal Variance Component Analysis showing the observed variance for each brain region individually over all sex-matched samples (3, 12, 15, 18, 21 months) for the biological factors age and sex. The point size indicates the variance when observing age and sex in combination. Combined variance shares are indicated in point size. The colors refer to the regions from A. Thresholds at 12% for both axes are marked with grey dashed lines. d Volcano plot for the sex-specific comparison of the brain regions mot. cortex, choroid plexus, thalamus and olfactory bulb. Colored dots indicate significantly down (green) and significantly upregulated (yellow) miRNAs (fold change ≥ 1.5 or ≤ 1/1.5, adjusted p-value < 0.05). e The gene set enrichment analysis (GSEA) result obtained from MIEAA for three brain regions showing the top 10 (sorted adjusted by p-value) depleted (green) and enriched (yellow) pathways over the three brain regions.

Fig. 3

Sex-specific miRNA dynamics in brain aging across male and female. Sex-specific analysis. We split the figure in a left and a right part covering the male and female samples, respectively. a Boxplots showing the time series of miRNA mmu-miR-9-5p for three different brain regions each. The Spearman’s rank correlation coefficient from the miRNA expression with age is displayed above each plot. b Bar plots showing the number of significantly anti- or positively correlated miRNAs with age per brain region (using Spearman’s rank correlation coefficient with |R| ≥ 0.5, adjusted p-value < 0.05). The upper bars contain miRNAs which are in the same direction significantly correlated in more than one brain region (in yellow). The respective lower bars contain miRNAs that are unique for one brain region (colored in the corresponding brain region color). c List of features which are significantly correlated for more than one brain region. The dots indicate the brain region(s) (colored according to Fig. 1a) in which the feature is significantly positively correlated (yellow; top) and significantly anti-correlated (green; bottom). d Heatmaps showing for each brain region the number of up- or deregulated miRNAs for the comparison between each age stage to 3 months (fold change ≥ 1.5 or ≤ 1/1.5, adjusted p-value < 0.05). Numbers in cell displayed when above threshold of 200 miRNAs. e List of features which are in the same direction significantly deregulated for at least one age comparison within at least two brain regions with significantly upregulated miRNAs at the top and significantly downregulated ones at the bottom. Analogous to Fig. 3c, the colors of the dots indicate the brain region in which the deregulations can be found. f Shows the amount of brain region-specific and age-related features per approach. Purple markers indicate the number of brain region-specific features from Fig. 2a, arrows pointing to the top (bottom) show the amount of significantly upregulated (downregulated) miRNAs in at least one age comparison and positive (negative) signs show the amount of significantly positively (anti-) correlated miRNAs. Human miRNA data from ROSMAP: g We divide the healthy samples into three similar sized sets according to their age of death (71 to 80, 81 to 91 and 92 to 102 years). Using the youngest and the oldest group we perform a DE analysis per sex. The color denotes the fold change for the feature (columns) and the black border if the fold change is greater or equal than 1.5 or smaller or equal than 1/1.5. A star would indicate a significance based on the adjusted p-value (Student’s t-test, Benjamini-Hochberg procedure). h Scatter plot of the Cohen’s d values against the log2-normalized fold changes colored if exceeding the thresholds for the fold change (|log2(fc)| ≥ log2(1.5)) and of the effect size (|d| ≥ 0.5). Yellow for an upregulation and green in case of a downregulation. The labeled miRNAs are those which were already identified as age-correlated for the brain region mot. cortex (Supplementary Fig. 4c).

Fig. 4:

Global miRNA expression dynamics in brain aging across regions. In this figure, we consider all samples from our data set. a On the right heatmaps showing the number of deregulated miRNAs for each brain region regarding the comparison between each age stage to 3 months (fold change ≥ 1.5 or ≤ 1/1.5). Numbers in cell displayed when above threshold of 200 miRNAs. On the left numbers of significantly anti- or positively correlated miRNAs with age per brain region (using Spearman’s rank correlation coefficient with |R| ≥ 0.5, adjusted p-value < 0.05). b Upset plot providing an overview of miRNAs changing with age per brain region and presenting the uniqueness or overlap of these candidates. To determine these candidates, we determine the Spearman’s rank correlation coefficient for each miRNA in each brain region with age and filter for significantly positively or anti-correlated features (|R| ≥ 0.5, adjusted p-value < 0.05). Additionally, we perform a DE analysis for all comparisons between each age stage to 3 months. Additional candidates are those features for a brain region if the feature is significantly deregulated (abs. log2-fold change ≥ 1.5, adjusted p-value < 0.05) for at least one comparison. We highlight a brain region in bold if it is above the threshold in Fig. 2c concerning a high variance within age. For simplicity of the plot, we do not display any brain region with less than 10 candidates and combinations with less than 5 miRNAs. c Overview of the results of a trajectory clustering. We clustered the standardized time series for each miRNA and each brain region in k=53 clusters with a membership of at least 15%. Each dot corresponds to one cluster and the size of the dot corresponds to the number of trajectories in the cluster. We determine which brain region is most frequent in each cluster (relative to the cardinality of each cluster) and show the occurrence on the x-axis. If the value exceeds a threshold of 30%, we color that dot according to the brain region (see Fig. 1a). The y-axis shows the incidence of the most frequent feature in each cluster. If the value exceeds 4, we display the dot in a dark green color. Else it is colored in light green. d The displayed center lines belong to the clusters having a brain region occurrence higher than 30%, highlighted are the center lines for clusters with choroid plexus and pons. e Trajectories for four chosen clusters. Each line corresponds to the z-scored miRNA expression for one brain region. The two left plots show a steady increase with age and the two right plots show a steady decrease. At the top, we list the most frequent brain regions (rel. to the cardinality of that cluster) and the number of the most frequent miRNAs if the occurrence exceeds 10% in the brain region case and if the occurrence exceeds 3 in the feature case. Brain regions with a black border or bold miRNAs exceed the thresholds introduced in Fig. 4c. f A more detailed distribution of brain region occurrence (left) and miRNA frequency (right) for each cluster which exceeds any threshold from Fig. 4c. For simplicity, we do not show brain regions with an occurrence < 10% and a miRNA incidence < 3. For incidences exceeding the threshold from Fig. 4c (grey dashed lines), we provide the miRNA names and color the markers in dark green. For a better discriminability, we added a black border to the dots corresponding to the brain regions over the threshold. g The center lines of seven clusters containing trajectories corresponding to mmu-miR-155-5p. If trajectories for multiple brain regions were assigned to the same cluster and therefore have the same center line, we vary the line width for visualization purposes. h Boxplots showing the trajectories of miRNA mmu-miR-155-5p for seven brain regions. Asterisks highlight the significance of a deregulated comparison between that age and the control age (3 months). The Spearman’s rank correlation coefficient from the miRNA expression with age is displayed above each plot (significantly, if adjusted p-value < 0.05).

Fig. 5:

MiR-155-5p expression and target interactions in brain aging and microglia. a Adjusted from CNS microRNA Profiles. We display the occurrences of mmu-miR-155-5p in different brain cell types via the fold changes between the cell type and the brainstem data. b Venn plot created in distinct mode for our microglia expression data derived from FACS sorted cells from young and aged mice compared to young microglia miRNA expression profiles derived from immunopanning. c The top 25 most expressed miRNAs in our microglia data set (calculated over all samples) depicted in a log10-scale. The miRNAs are clustered using hierarchical clustering. d Top 25 miRNAs according to the coefficient of variation. We show the absolute standardized expression for each of the miRNAs and samples. Both the miRNAs and the samples are clustered via hierarchical clustering. e Line plot for two miRNAs showing the median values for young and old indicating the direction of deregulation and the expression. f Volcano plot presenting the DE analysis of the comparison old versus young showing the raw p-values. No miRNA is significantly deregulated considering the adjusted p-values. Nonetheless, we observe a strong upregulation for mmu-miR-155-5p marked in bold. Aging cohort: g Trajectories over the medians per age showing the expression of different isomiRs of mmu-miR-155 in five different brain regions. The canonical isomiR is highlighted as a dark line. h We calculate the 25 most expressed isomiRs decreasing from top to bottom over all samples. The averaged standardized expression values per brain region and isomiR are clustered according to the brain regions (using hierarchical clustering with complete linkage) and are split into three clusters. i Region-wise correlation values (Spearman’s rank correlation coefficient) for target genes (obtained from miRTargetLink 2.0) of mmu-miR-155-5p using the mRNA data released by Hahn et al.. The upset plot only shows the significantly anti-correlated genes ((|R| ≥ 0.3, adjusted p-value < 0.05) and the brain region in which the gene is significantly anti-correlated. Genes are colored by pathways: “mTOR Signaling Pathway” (yellow) and the “Regulation of cell communication” (blue). j The scatter plots are given for four of the significantly anti-correlated target genes shown in Fig. 5i. We see in each plot the brain regions in which this gene is significantly anti-correlated (indicated by the color) and the relation between the gene and the miR-155-5p broken down into median per brain region and age point.