Plasma microRNA signature associated with retinopathy in patients with type 2 diabetes

Screening for differentially expressed circulating miRNAsFor an initial screening, the expression of over 170 plasma-enriched miRNAs was assessed on samples obtained by pooling RNAs extracted from plasma exosomes of diabetic individuals without DR (Control) and with DR (Retinopathy). Type 2 diabetes is associated with profound changes in circulating (and exosomal) miRNAs8,12,13, hence an additional control group including individuals without diabetes and retinopathy was included in order to assess differences in miRNAs expression compared to the general population as well. Overall, 22.3% of the tested miRNAs were detected in all the pooled samples and we identified 18 circulating miRNAs potentially dysregulated (FC ≥ |8.0|) in patients with DR compared with diabetic individuals (Fig. 1a). Conversely, comparison of patients with DR with non-diabetic individuals showed dysregulation of 19 miRNAs (Fig. 1b). Among them, 12 miRNAs (let-7a-5p, miR-16-5p, miR-23a-3p, miR-25a-3p, miR-27a-3p, miR-92a-3p, miR-150-5p, miR-197-3p, miR-223-3p, miR-320a-3p, miR-320b, miR-486-5p) were upregulated and 2 miRNAs (miR-346 and miR-495-3p) showed a consistent downregulation in both comparisons (Fig. 1c). These 14 miRNAs and were selected for the following validation step.Figure 1Circulating miRNome screening of DR. (a,b) Analysis of circulating miRNAs in pooled samples of diabetic patients with DR (Retinopathy) vs. diabetic patients without DR (Control) (a) or individuals without diabetes and retinopathy (No DM) (b). The x-axis indicates the fold changes on a log2 scale. DM, diabetes mellitus. (c) Venn’s diagram depicts the proportion of miRNAs consistently dysregulated in analyses in (a) and (b). The list of miRNAs is reported below.Internal validation of differential regulationAs analyses of pooled samples could not provide definitive results in terms of statistical significance, the expression of the selected miRNAs was measured in diabetic patients with and without DR. For this analysis, additional 10 diabetic patients with DR were included to enhance statistical power. Among the tested miRNAs, we found a significant upregulation of miR-23a-3p (log2FC = 2.65 ± 0.92, P = 0.005), miR-25-3p (log2FC = 3.54 ± 1.13, P = 0.004), and miR-320b (log2FC = 2.53 ± 0.93, P = 0.011), while miR-495-3p was dramatically downregulated (log2FC = − 4.21 ± 0.97, P < 0.001) (Fig. 2a,b). Of note, these differences remained significant also after accounting for multiple comparisons according to Benjamini–Hochberg FDR, while other miRNAs (miR-92a-3p and miR-346) were not confirmed as significant (Table S1). The expression of these 4 miRNAs was then compared to non-diabetic individuals: while expression of miR-25-3p (log2FC = 2.75 ± 0.99, P = 0.004), miR-320b (log2FC = 2.65 ± 0.54, P < 0.001), and miR-495-3p (log2FC = -3.25 ± 0.89, P = 0.003) were significantly different, we did not detect significant differences in miR-23a-3p expression (log2FC = 1.83 ± 0.91, P = n.s.) between patients with DR and non-diabetic subjects (Fig. 2c).Figure 2Validation of miRNA regulation by qPCR. (a,b) Expression of the selected miRNAs was performed by qPCR on diabetic patients without DR (Controls, n = 10) and with DR (Retinopathy, n = 20). Heatmap shows the mean expression of the miRNA in each group (a) and the Volcano plot reports the P-values by multiple t-tests (b). (c) Expression of the miRNAs showing a significant dysregulation in (b) measured in non-diabetic subjects (n = 10), diabetic patients without (n = 10) and with DR (n = 20). T2DM, type 2 diabetes. *p < 0.05; **p < 0.01; ***p < 0.001 as computed by univariate ANOVA with Bonferroni post-hoc test.Association of miRNAs with severity of DRNext, we investigate the association between circulating miRNAs expression and DR severity, staged according to the International Clinical Disease Severity Scale for DR. Interestingly, significant moderate grade correlations between severity of DR and circulating miR-25-3p (Spearman’s ρ = 0.52, P = 0.001), miR-320b (ρ = 0.51, P = 0.001), and miR-495-3p (ρ = − 0.46, P = 0.003) were found, showing a linear trend in regression analyses (Fig. 3a) which was independent on age and gender (Table S2). Conversely, no significant correlation was observed for miR-23a-3p (ρ = 0.28, P = n.s.). Moreover, logistic regression analyses confirmed the significant associations of circulating miR-25-3p (Odds ratio, OR = 2.50, 95% confidence interval = 1.38–4.52, P = 0.003), miR-320b (OR = 2.29, 1.30–4.03, P = 0.004), and miR-495-3p (OR = 0.40, 0.22–0.73, P = 0.003) with DR, that persisted after adjustment for age, gender, and HbA1c in multivariate logistic regression models (Fig. 3b, Table S3). Furthermore, no significant correlations were detected between circulating miR-23a-3p, miR-25-3p, miR-320b, miR-495-3p, and age, nor these miRNAs displayed sex-specific regulation (Fig. S1, Table S4). Finally, since combination of multiple miRNAs could yield superior discriminative performance than individual miRNAs, we conducted logistic regression analysis employing the circulating level of miR-25-3p, miR-320b, and miR-495-3p as independent variables. The overall model resulted statistically significant (Cox and Snell R2 = 0.526, P < 0.001) and the derived Receiver Operating Characteristic (ROC) curve demonstrated a good accuracy of this model for statistically detecting DR in our study cohort (AUC = 0.931, 0.853–1.000, sensitivity and specificity 85%, P < 0.001) (Fig. 3c) and correlated with DR severity with larger coefficients (ρ = 0.79, P < 0.001) than individual miRNAs.Figure 3Circulating miR-25-3p, miR-320b and miR-495-3p are associated with DR. (a) Univariate linear regression models showing log-linear trends of association between miRNAs and severity of DR. Grade I, no retinopathy (n = 10); Grade II, mild no-proliferative DR (NPDR, n = 5); Grade III, moderate NPDR (n = 5); Grade IV, severe NPDR (n = 7); Grade V, proliferative DR (n = 3). (b) Forest plot showing odds ratios and the 95% CI for association of the single miRNAs with presence of DR in univariate logistic models (upper) or after correction for age, sex, and HbA1c (lower). (c), Receiver-operating characteristic (ROC) curve derived from the logistic regression model developed on circulating miR-25-3p, miR-320b, and miR-495-3p. AUC, area under the curve.Lastly, we explored the association between DME, a relevant cause of vision loss that could occur at any stage of DR1, and dysregulated miRNAs. Logistic regression analyses revealed significant associations for miR-320b (OR = 2.41, 1.27–4.56, P = 0.007) and miR-495-3p (OR = 0.46, 0.23–0.94, P = 0.035). Entering both miRNAs as covariates yielded a statistically significant model (Cox and Snell R2 = 0.287, P = 0.001) with a good diagnostic performance as revealed by ROC curve analysis (AUC = 0.847, 0.722–0.972, sensitivity 83%, specificity 79%, P < 0.001).Prediction of biological relevanceCirculating miRNAs in exosomes have been proposed as a paracrine (and potentially endocrine) intercellular communication system6,7,8,9,10. Hence, we explored the network of the biochemical interactions involving the validated targets of the dysregulated miRNAs. The list of miRNA targets with functional validation is reported in Table S5. Intriguingly, a low target overlap was found as only miR-23a-3p and miR-25-3p shared two common targets (CDH1 and PTEN) (Fig. 4a). The complete network of biochemical interactions is shown in Fig. 4b. Of note, gene ontology analysis revealed these targets to be involved in crucial pathways including regulation of metabolic process, regulation of cell response to stress, and development of blood vessels (Fig. 4c). Moreover, mapping the targets to the REACTOME database identified a significant enrichment in processes relevant for endothelial cell proliferation20, such as the regulation of the NOTCH1 (e.g. R-HSA-4641262, Fold enrichment = 11.5, P = 0.027) and the β-catenin pathways (e.g. R-HSA-2559580, Fold enrichment = 4.9, P = 0.018). Together, these data suggest a possible biological relevance of the dysregulated miRNAs in influencing (endothelial) cell proliferation and—possibly—angiogenesis.Figure 4Biological pathway network. (a) Venn’s diagram for the overlap of target mRNA for the dysregulated miRNAs. (b) Network analysis of biochemical interaction among the validated targets of the dysregulated miRNAs as enlisted in the STRING database. (c) Gene ontology analysis for enrichment in biological processes pathways of the validated targets of the dysregulated miRNAs.

Via Source link