Coxsackievirus B5 Infection Induces Dysregulation of microRNAs Predicted to Target Known Type 1 Diabetes Risk Genes in Human Pancreatic Islets

Type 1 diabetes is an immune-mediated disease resulting from the complex interplay between genetic and environmental factors. The role of enteroviruses (EVs) as key triggers of this disease has been heavily debated due to the lack of unequivocal proof that EVs serve as causal agents, despite significant associations between EV infection and islet autoimmunity (odds ratio 3.7) and clinical type 1 diabetes (odds ratio 9.8) (1). Among EV strains implicated with this disease, coxsackievirus B (CVB) group has been studied most extensively. Multiple mechanisms by which CVB might initiate and/or accelerate autoimmunity have been proposed, including direct induction of β-cell death, promotion of inflammatory cytokine production, activation of toll-like receptors (TLRs), molecular mimicry, and most recently, dysregulation of host microRNAs (miRNAs) (2,3).

Sized between 20 and 22 nucleotides, miRNAs are noncoding RNAs that function as potent regulators of gene expression (4). Derived from sequential processing of longer precursor transcripts, they load Argonaute proteins to mediate either destabilization or translational inhibition of target mRNAs. Sequence specificity of this control is provided by the complementary base pairing between the miRNA and mRNA, usually within its 3′ untranslated region (3′UTR).

Growing evidence supports pivotal roles of miRNAs in β-cell biology, including regulation of differentiation, glucose sensing, insulin exocytosis, inflammation, and apoptosis (3,5–7). Moreover, a number of miRNAs have been identified as components of pathways initiated by or contributing to the pathogenesis of both type 1 and type 2 diabetes (3). Therefore, it can be anticipated that changes in miRNA activities elicited by environmental exposures contribute significantly to the initiation and acceleration of islet autoimmunity.

To date, very few studies have examined global miRNA expression in the pancreas and none in CVB-infected human β-cells. In fact, only one study has investigated the effect of CVB infection on pancreatic miRNA expression, using rat β-cells (8). Therefore, the aim of this study was to identify miRNAs dysregulated in CVB5-infected human pancreatic islets and investigate whether they target genes associated with risk of type 1 diabetes.

Human cadaveric pancreatic islets from two anonymous donors were cultured separately and infected with one clinical CVB5 strain (GenBank accession no. GQ126860.1) at 1 multiplicity of infection (MOI), as previously described (9). MOI was optimized to ensure >95% viability of cells in culture after viral exposure using the trypan blue dye exclusion test. Cells were harvested at 1, 4, and 7 days postinfection (dpi) and stored in TRI Reagent (Ambion) at −80°C.

RNA extraction, cDNA synthesis, and preamplification were performed as previously described (10). Human Pool A v2.1 and B v3.0 Megaplex stem-loop RT Primers (Applied Biosystems) were used to produce cDNA. After preamplification, cDNA was diluted (1:40), combined with an equal volume of Master Mix, and loaded onto the TaqMan OpenArray Real-Time PCR Plates containing 754 miRNA assays. Plates were analyzed on the QuantStudio 12k Flex Real-Time PCR System using recommended settings. For each batch of human islets, experiments were repeated at least three times and evaluated in triplicate.

Total RNA (1 μg) from control and CVB5-infected cells (4 dpi) were DNase treated (Promega) before cDNA synthesis using Superscript III (Invitrogen) and random hexamers. Resulting cDNA was diluted equally (1:10) before quantitative PCR, performed using SYBR FAST Master Mix (Kapa Biosystems) and gene-specific primers (5′-3′): BACH2 (GCCTCAATGACCAGCGGAAA; CAAACAGGCCATCCTCACTG), GAPDH (TCAAGATCATCAGCAATGCCTCC; ATCACGCCACAGTTTCCCG), β-ACTIN (CTGTACGCCAACACAGTGCT; GCTCAGGAGGAGCAATGATC), SH2B3 (AACCACCAGGTTCCTGCAAC; GGACAGCCAGAAGAACTAAGGTG), GLIS3 (CAACCAGATCAGTCCTAGCTTACA; GCGAAATAAGGGACCTGGTATC), CLEC16A (CATCAAGACGAGTGGGGAGAGT; TCCTCGTCCGTGGTGTTCTG), TLR7 (CCAGATATAGGATCACTCCATGCC; CAGTGTCCACATTGGAAACACC), and IKZF1 (CACAGTGAAATGGCAGAAGACC; GGCCCTTGTCCCCAAGAAAT).

The following cycle was repeated 40 times on the LightCycler 480II (Roche): 95°C 10 s, 60°C 20 s, and 72°C 15 s. Assays were performed in triplicate and confirmed at least three times. GAPDH and β-ACTIN levels were used to normalize Ct values. Relative mRNA expression for each gene was expressed as a fold difference to control expression, and significance was evaluated using Student t tests.

Only miRNAs with Ct values ≤35 and amplification scores ≥1.15 were considered for normalization. Global mean normalization (11) was applied to compare the geometric mean of all miRNA Ct values to individual miRNAs using ExpressionSuite v1.0.3 (Life Technologies). Fold differences between control and infected cells were calculated using the 2^−ΔΔCt method (12). When fold difference was <1, the negative reciprocal was taken to express a negative fold change. Only miRNAs with ≥ threefold difference were included for target prediction analysis using miRWalk (13). All identified miRNAs and type 1 diabetes risk genes compiled from T1DBase (14) were assessed by the miRWalk predictor tool using all 10 databases with the following settings: “human species,” “restricted to 3′UTR,” “2,000 upstream flanked,” “transcript = longest,” “minimum seed length = 7,” and “pValue = 0.05.” Genes predicted as targets by three or more databases were functionally annotated using the Database for Annotation, Visualization, and Integrated Discovery (DAVID), version 6.7 (15), which clustered them according to their associated gene ontology annotations and assigned each cluster an enrichment score (−log[P value]).

The nonparametric Mann-Whitney U test was used to compare normalized Ct values of miRNAs in CVB5-infected versus control cells. Significance was set at an unadjusted P value <0.05 and miRNAs satisfying this were included for further analyses. For DAVID, biological processes with a false discovery rate of ≥0.05 computed by the Benjamini-Hochberg correction were deemed not statistically significant and discarded. Statistical analyses were performed using Prism (version 6; GraphPad Software Inc., San Diego, CA) and SPSS (version 22; IBM, Armonk, NY).

After CVB5 infection, two miRNAs were significantly dysregulated (≥ threefold difference compared with mock-infected control) in human pancreatic islets at 1 dpi, with miR-155-5p increased fourfold and miR-181a-3p decreased threefold (Table 1). In contrast, 19 miRNAs were significantly decreased at 4 dpi (Fig. 1A). At 7 dpi, 12 miRNAs were decreased and 5 were increased (Fig. 1B). Overall, levels of 33 miRNAs were significantly altered during 7 days of infection, with 6 upregulated and 27 downregulated (Table 1). Greatest upward and downward fold changes were exhibited by miR-34a-5p and miR-625-5p, respectively. All differences measured between control and CVB5-infected islets were statistically significant (P < 0.05). Of the 33 CVB5-responsive miRNAs, 6 ranked among the top 100 most abundant miRNAs expressed in purified β-cells of healthy humans (7).

To explore the functional significance of these dysregulations, miRWalk database was used for miRNA target prediction (13). Alignments were restricted to the 3′UTRs of all candidate type 1 diabetes risk genes reported at the time of investigation (14). Remarkably, 57 of the total 72 risk genes were predicted as targets (Table 2). More specifically, miRNAs dysregulated at 1, 4, and 7 dpi were associated with 8, 56, and 47 different target risk genes, respectively. Most miRNAs were predicted to target multiple genes, with many exhibiting significant overlap. BACH2 displayed the greatest overlap, with 21 miRNAs predicted to target its 3′UTR (Table 2).

Biological processes of the 57 putative targets were investigated using DAVID (15), which grouped them according to their respective gene ontology annotation and provided an enrichment score for each cluster (Fig. 2A). No statistically significant biological processes were identified from the predicted targets of miRNAs dysregulated at 1 dpi. In contrast, 67 common processes were identified from the predicted targets of miRNAs differentially expressed at 4 and 7 dpi. The process with the highest enrichment score was “positive regulation of immune processes.”

CVB5 infection can induce upregulation of many immune response genes in human islets, including pattern recognition receptors (PRRs) (16). Indeed, mRNA levels of IFIH1 and TLR7 were significantly elevated in the CVB5-infected cells compared with controls at 4 dpi (Fig. 2B). Examination of other type 1 diabetes risk genes revealed similar increases in the mRNAs of SH2B3 and IKZF1, whereas levels of BACH2, GLIS3, and CLEC16A remained unchanged.

Our examination of CVB5-infected human islets revealed significant dysregulation of 33 miRNAs during 7 days of infection (Table 1). This represents 2.2% of all characterized human miRNAs (5) and 6.4% of miRNAs expressed in healthy human pancreatic β-cells (7). Since a single miRNA can potentially regulate hundreds of genes (4), disruption of even a small subset of miRNAs can profoundly impact the transcriptome and biology of cells.

Both miR-155-5p and -181a-3p are important modulators of the immune response, directing suppression of proinflammatory cytokines and NF-kB pathway activators (17,18). In this study, miR-155-5p was elevated in CVB5-infected cells and predicted to target the antiapoptotic gene GLIS3. Loss of GLIS3 sensitizes cells to cytokine-mediated apoptosis (19). Therefore, increased miR-155-5p may reduce GLIS3 expression and promote β-cell apoptosis, whereas the concurrent reduction of miR-181a-3p is expected to produce the opposite effect. Although it seems contradictory that two miRNAs negate each other’s effects, apoptosis is highly disadvantageous for the virus, and therefore CVB5 may directly decrease miR-181a-3p to nullify the host’s promotion of apoptosis via miR-155-5p.

Among miRNAs dysregulated at later time points, miR-10b-3p, -182-5p, -186-5p, -345-5p, -376c-3p, -425-3p, -432-5p, -493-3p, and -629-3p were predicted to target multiple PRRs: IFIH1, TLR7, and TLR8 (Table 2). These encode proteins that promote generation of type 1 interferons and proinflammatory cytokines (20). They also activate downstream caspases that trigger apoptosis (21). All PRR-targeting miRNAs were downregulated in response to CVB5 infection, except miR-10b-3p. Therefore, the likely outcomes are the overexpression of PRRs and enhanced immune response against CVB5. Although this seems beneficial for viral defense, such elevation could induce exacerbated inflammation and prolonged apoptosis of β-cells, promoting type 1 diabetes development.

BACH2 is an important regulator of cytokine-induced apoptosis in pancreatic β-cells (22). In rodent and human pancreatic islets, BACH2 inhibition exacerbates cytokine-induced β-cell destruction. Conversely, its overexpression has protective effects. Therefore, it is plausible that the reduction of miRNAs targeting BACH2 at 4 and 7 dpi results in BACH2 overexpression, protecting CVB5-infected cells from apoptosis. This could be a strategy imposed by CVB5 to promote its propagation. If so, the increase of several BACH2-targeting miRNAs observed at 7 dpi (Table 2) might reflect the host’s efforts to counter CVB5 and restore normal apoptosis of infected β-cells.

Interleukin-2 (IL-2), IL-10, IL-17D, and IL-21 were among the predicted targets of miRNAs downregulated in CVB5-infected cells. Reduced miRNA-directed suppression of these cytokines will likely enhance inflammation and β-cell death, potentiating islet autoimmunity. Consistent with this hypothesis, we previously identified significant correlations between increased levels of proinflammatory cytokines (including IL-2, IL-17, and IL-21) and islet autoimmunity in children (23). Furthermore, expression of many type 1 diabetes risk genes are modified by cytokine exposure in human pancreatic islets, including PTPN2, IFIH1, SH2B3, STAT-4, GLIS-3, CD55, RASGRP1, and SKAP2 (24), which are all predicted targets of CVB5-dysregulated miRNAs identified in this study.

The greatest fold increase was exhibited by miR-34a-5p (eightfold) (Table 1). Previously, its expression increased three- to fourfold after exposure to palmitate and proinflammatory cytokines in rodent and human pancreatic islets (6,25). Moreover, similar elevation of miR-34a was present in β-cells of db/db and NOD mice during development of prediabetic insulitis (6,25). Two targets of miR-34a-5p have been verified:

1. VAMP2, essential for β-cell exocytosis, and

2. antiapoptotic BclII.

Accordingly, miR-34a-5p overexpression reduces both targets in MIN6 cells, diminishing maximum insulin secretion capacity and increasing apoptosis (25). Thus, the sharp increase of miR-34a-5p in CVB5-infected islets may enhance secretory dysfunction and destruction of β-cells, which are hallmarks of type 1 diabetes.

When interpreting fold changes in miRNA expression between test and control samples, it is crucial to consider the natural miRNA abundance. For example, miR-625-5p exhibited the greatest fold change between control and CVB5-infected cells. However, the 14-fold reduction of this miRNA is unlikely to impact β-cell function, as its normal expression in pancreatic β-cells is already very low (Table 1) (7). This also applies to miR-216b-5p, -376c-3p, -181a-3p, and -629-3p. In contrast, sudden increases of low-abundant miRNAs can lead to new miRNA-target interactions and regulatory cascades, which could be the case for miR-34a-5p, -21-3p, -155-5p, -1290, -663b, and -10b-3p.

Six miRNA-target interactions identified in this study (BACH2-miR-29a-3p, BACH2-miR-34a-5p, SH2B3-miR-193b-3p, TNFAIP3-miR-29a-3p, ZFP36L1-29a-3p, and RGS1-miR-29a-3p) have been validated in pancreatic β-cells of healthy adults (7). However, others require future experimental verification. To fully evaluate the significance of our findings, it is important to determine whether the dysregulation of 33 miRNAs in CVB5-infected human islets produces concomitant changes in protein abundances of predicted targets. One high-throughput approach would be to compare global peptide profiles of CVB5-infected and uninfected human pancreatic islet cells using proteomics. In this study, we measured the mRNA levels of seven predicted targets and found significant elevations for IFIH1, TLR7, SH2B3, and IKZF1 in CVB5-infected islets (Fig. 2B). This is consistent with the decreased levels of miRNAs targeting these genes, which could subsequently reduce the destabilization of their mRNAs. In contrast, genes that exhibited no change at the mRNA level may only be affected at the protein level.

In summary, our study represents the first global miRNA expression profiling of CVB5-infected human pancreatic islets. Collectively, the 33 dysregulated miRNAs identified were predicted to target mRNAs of 57 candidate type 1 diabetes risk genes. Given that the majority of these miRNAs were significantly downregulated, and their targets function as key regulators of the proinflammatory response, it is plausible that the disrupted activities of these miRNAs induced by CVB5 may exacerbate the antiviral immune response in genetically susceptible individuals, initiating a pathway that leads to the destruction of β-cells and eventually type 1 diabetes.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. K.W.K. researched data and wrote the manuscript. A.H. and A.A.H. researched data and contributed to discussion. A.A.-A. researched data. T.W.H.K. provided the human pancreatic islets used in this study. W.D.R. and M.E.C. researched data, contributed to discussion, and reviewed and edited the manuscript. M.E.C. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.