Candidate master microRNA regulator of arsenic-induced pancreatic beta cell impairment revealed by multi-omics analysis

Abstract

Arsenic is a pervasive environmental toxin that is listed as the top priority for investigation by the Agency for Toxic Substance and Disease Registry. While chronic exposure to arsenic is associated with type 2 diabetes (T2D), the underlying mechanisms are largely unknown. We have recently demonstrated that arsenic treatment of INS-1 832/13 pancreatic beta cells impairs glucose-stimulated insulin secretion (GSIS), a T2D hallmark. We have also shown that arsenic alters the microRNA profile of beta cells. MicroRNAs have a well-established post-transcriptional regulatory role in both normal beta cell function and T2D pathogenesis. We hypothesized that there are microRNA master regulators that shape beta cell gene expression in pathways pertinent to GSIS after exposure to arsenicals. To test this hypothesis, we first treated INS-1 832/13 beta cells with either inorganic arsenic (iAs^III) or monomethylarsenite (MAs^III) and confirmed GSIS impairment. We then performed multi-omic analysis using chromatin run-on sequencing, RNA-sequencing, and small RNA-sequencing to define profiles of transcription, gene expression, and microRNAs, respectively. Integrating across these data sets, we first showed that genes downregulated by iAs^III treatment are enriched in insulin secretion and T2D pathways, whereas genes downregulated by MAs^III treatment are enriched in cell cycle and critical beta cell maintenance factors. We also defined the genes that are subject primarily to post-transcriptional control in response to arsenicals and demonstrated that miR-29a is the top candidate master regulator of these genes. Our results highlight the importance of microRNAs in arsenical-induced beta cell dysfunction and reveal both shared and unique mechanisms between iAs^III and MAs^III.

Keywords: Arsenic; Beta cells; Diabetes; Insulin secretion; MicroRNAs; Sequencing.

Fig. 1

Exposure to arsenicals impairs glucose stimulated insulin secretion in INS-1 832/13 cells. INS-1 832/13 cells were exposed to 0.5 μM of MAs^III, 1 μM of iAs^III (A), or 2 μM of iAs^III (B) for 24 h. Secreted insulin was quantified via insulin ELISA assays in technical and biological replicates of 3 and normalized to INS-1 832/13 cells incubated with high glucose and no arsenical. Experimental run 1 (A) and run 2 (B) were performed by two separate technicians with separately prepared arsenicals. Two-tailed unpaired Students t test was used to calculate the p values: *p value < 0.05, **p value < 0.01 arsenical treatment versus untreated high glucose control

Fig. 2

Exposure to iAs^III and MAs^III alters miRNA expression in INS-1 832/13 cells similarly. A Principal components analysis (PCA) plot of miRNA profiles in iAs^III treated cells compared to control group (untreated). Batch 1 and 2 were performed by separate technicians and therefore batch correction using the limma package was implemented. PCA plot was generated after variance stabilizing transformation (VST) of the data. B Volcano plot representing the differential expression (DE) analysis of miRNAs after iAs^III treatment. 10 significantly altered miRNAs are highlighted: 6 downregulated genes and 4 upregulated (p adjusted < 0.05, log2fold-change < − 0.5 or > 0.5, basemean > 500). C PCA plot of miRNA profiles in MAs^III treated cells compared to control group (untreated). PCA was generated after applying VST. All MAs^III treatments were performed by a single technician. D Volcano plot representing the DE analysis of miRNAs after MAs^III treatment. 4 significantly altered miRNAs are highlighted: 1 downregulated and 3 upregulated, in the MAs treatment compared to the control group (p adjusted < 0.05, log2fold-change < − 0.5 or > 0.5, basemean > 500). E Venn diagrams showing shared and uniquely altered miRNAs between iAs and MAs. F Significantly altered miRNA expression in iAs and MAs^III treatment groups. Normalized counts generated by DESeq2. The Benjamini–Hochberg method was used to calculate the adjusted p value: **p adjusted < 0.01

Fig. 3

Exposure to iAs^III or MAs^III leads to unique changes in gene expression profiles in INS-1 832/13 cells. A PCA plot of gene expression profiles in iAs^III and MAs^III treated cells compared to the control group (untreated). Batch 1 and 2 were performed by separate technicians and therefore batch correction using the limma package was implemented. PCA plot was generated after applying VST. B Volcano plots representing the DE analysis of genes after iAs^III or MAs^III treatment (significance indicated by red or blue; p adjusted < 0.05, log2fold-change < − 0.5 or > 0.5, basemean > 500). C Venn diagrams showing shared and uniquely altered genes between iAs^III and MAs^III. D Significantly altered gene expression in iAs^III and MAs^III treatment groups. Normalized counts generated from DESeq2 analysis. The Benjamini–Hochberg method was used to calculate the adjusted p value: **p adjusted < 0.01

Fig. 4

Exposure to iAs^III or MAs^III leads to unique changes in gene transcription profiles in INS-1 832/13 cells. A Correlation analysis between gene fold-changes at the RNA-seq level and ChRO-seq level (iAs^III: R = 0.45, p value < 2.2e − 16; MAs^III: R = 0.57, p value < 2.2e − 16). Only genes with baseMean > 500 in the RNA-seq data are shown. B PCA plot of altered gene transcription profiles in iAs^III and MAs^III treated cells compared to the control group (untreated). The limma package was used for batch correction between experimental runs. PCA plot was generated after applying VST. C Volcano plots representing differentially transcribed (DT) genes after iAs^III or MAs^III treatment (significance indicated by red or blue; p adjusted < 0.05, log2fold-change < − 0.5 or > 0.5). D Venn diagrams showing shared and uniquely altered genes between iAs^III and MAs^III. E Significantly altered genes both transcriptionally and at the mRNA level in the iAs^III treatment group. F Significantly altered genes at only the mRNA level in the iAs^III treatment group. G Significantly altered genes both transcriptionally and at the mRNA level in the MAs^III treatment group. H Significantly altered genes at only the mRNA level in the MAs^III treatment group. Normalized counts generated from DESeq2 analysis. The Wald test was used to calculate p values: *p value < 0.05, **p value < 0.01

Fig. 5

miR-29a is a candidate master regulator for arsenical-induced post-transcriptional changes in gene expression. A Two-factor analysis reveals genes significantly altered according to RNA-seq (p adjusted < 0.05, log2fold-change < − 0.5 or > 0.5) but not significantly altered according to ChRO-seq (p adjusted > 0.2) for iAs^III (top) and MAs^III (bottom). GPS genes are in purple and LPS genes are in orange. B Percentage of GPS genes that are miR-29a predicted targets in rats is shown. C Venn diagram representing overlap in conserved (between rat and two other species among human, mouse, and dog) miR-29a target GPS genes in iAs^III and MAs^III treatment conditions. D Significantly altered miR-29a target GPS genes in iAs^III only, MAs only, or both treatment groups. Normalized counts generated from DESeq2 analysis. The Benjamini–Hochberg method was used to calculate the adjusted p values: *p adjusted < 0.05, **p adjusted < 0.01