Arsenic responsive microRNAs in vivo and their potential involvement in arsenic-induced oxidative stress

Abstract

Arsenic exposure is postulated to modify microRNA (miRNA) expression, leading to changes of gene expression and toxicities, but studies relating the responses of miRNAs to arsenic exposure are lacking, especially with respect to in vivo studies. We utilized high-throughput sequencing technology and generated miRNA expression profiles of liver tissues from Sprague Dawley (SD) rats exposed to various concentrations of sodium arsenite (0, 0.1, 1, 10 and 100mg/L) for 60days. Unsupervised hierarchical clustering analysis of the miRNA expression profiles clustered the SD rats into different groups based on the arsenic exposure status, indicating a highly significant association between arsenic exposure and cluster membership (p-value of 0.0012). Multiple miRNA expressions were altered by arsenic in an exposure concentration-dependent manner. Among the identified arsenic-responsive miRNAs, several are predicted to target Nfe2l2-regulated antioxidant genes, including glutamate-cysteine ligase (GCL) catalytic subunit (GCLC) and modifier subunit (GCLM) which are involved in glutathione (GSH) synthesis. Exposure to low concentrations of arsenic increased mRNA expression for Gclc and Gclm, while high concentrations significantly reduced their expression, which were correlated to changes in hepatic GCL activity and GSH level. Moreover, our data suggested that other mechanisms, e.g., miRNAs, rather than Nfe2l2-signaling pathway, could be involved in the regulation of mRNA expression of Gclc and Gclm post-arsenic exposure in vivo. Together, our findings show that arsenic exposure disrupts the genome-wide expression of miRNAs in vivo, which could lead to the biological consequence, such as an altered balance of antioxidant defense and oxidative stress.

Keywords: Arsenic; Glutamate–cysteine ligase; Glutathione; High-throughput sequencing; MicroRNA (miRNA); Oxidative stress.

Figure 1. Summary of Hi-Seq sequencing results

Total, annotated and precise matched reads of each sample were summarized. The ranges of total reads were from 3,000,000 to 16,000,000, and the annotated reads varied from 3.9% to more than 60% of total reads.

Figure 2. Unsupervised hierarchical clustering of miRNA expression data from liver tissue of 15 rats exposed to different concentrations of sodium arsenite

The miRNA profiles of 15 samples were clustered using the Manhattan distance metric in conjunction with Ward’s minimum variance clustering method. The 5 rectangle overlays indicate the membership of the 5 clusters when the dendogram is cut to that size. The gray scale bar at the base of the figure indicates the arsenic exposure for each of the samples. The association between iAs exposure and cluster membership is statistically significant (Fisher’s exact test, P-value=0.0012), which strongly supports the hypothesis that the steady state pattern of miRNA expression in the liver tissue can be modified by arsenic exposure.

Figure 3. Principle components analysis (PCA) of miRNA expression data

The miRNA profiles of 15 samples were subjected to a PCA analysis. The 15 samples were plotted according to their mappings into the space defined by the first two principal components (Upper panel). Sample RNA-083 was considered to possibly be an outlier so the PCA was repeated with respect to 14 sample miRNA profiles, excluding RNA-083 (Lower panel). The first two principal components accounted for 46% and 49% of the overall variability in the full set of 15 samples and the subset excluding RNA-083, respectively. When the profiles of miRNAs were projected into the two dimensional space defined by the first two principal components, the portion of the variability in the RNA-seq expression profiles that can be attributed to iAs exposure levels is significant enough to have observed weak clustering of the profile projections of samples with similar exposures.

Figure 4. RNA-seq and qPCR results for miR-151 and miR-183

The first column contains the RNA-seq normalized expression values as a function of iAs concentration while the second column contains the qPCR fold change estimates as a function of the same concentration levels. Both miR-151 and miR-183 were verified by qPCR at a nominal level of 0.05. The Jonkheere’s trend test p-values are listed in the figure subheadings for all plots. The Benjamini and Hochberg q-values are listed for the RNA-seq data. The data were expressed as mean ± SD (n=3). Oneway ANOVA were applied to compare the differences in expression level between various arsenic concentration groups and control group (*, P <0.05; **, P < 0.01). The P values were from ANOVA results.

Figure 5. RNA-seq and qPCR results for miR-148b and miR-192

The first column contains the RNA-seq normalized expression values as a function of iAs concentration while the second column contains the qPCR fold change estimates as a function of the same concentration levels. Neither miR-148b nor miR-192 was verified by qPCR at a nominal level of 0.05.The Jonkheere’s trend test p-values are listed in the figure subheadings for all plots. The Benjamini and Hochberg q-values are listed for the RNA-seq data. The data were expressed as mean ± SD (n=3). One-way ANOVA were applied to compare the differences in expression level between various arsenic concentration groups and control group (*, P <0.05; **, P < 0.01). The P values were from ANOVA results.

Figure 6. RNA-seq and qPCR results for miR-26a and miR-423

The first column contains the RNA-seq normalized expression values as a function of iAs concentration while the second column contains the qPCR fold change estimates as a function of the same concentration levels. Both miR-26a and miR-423 were verified by qPCR at a nominal level of 0.05. The Jonkheere’s trend test p-values are listed in the figure subheadings for all plots. The Benjamini and Hochberg q-values are listed for the RNA-seq data. The data were expressed as mean ± SD (n=3). One-way ANOVA were applied to compare the differences in expression level between various arsenic concentration groups and control group (*, P <0.05; **, P < 0.01). The P values were from ANOVA results.

Figure 7. Responses of hepatic GCL activity and GSH level to sub-chronic arsenic exposure in rats

Optimized experimental conditions were applied. Three rats were used for each concentration point. Low concentration (0.1 mg/L) arsenic exposure significantly increased GCL activities, whereas higher concentration treatment (10 and 100 mg/L) depleted GCL activities compared to the control. Hepatic GSH levels were significantly reduced with high concentration exposure (10 and 100 mg/L). * P<0.05; ** P<0.01

Figure 8. Expression of Nfe2l2, Gclc, Gclm, Keap1 and Bach1 mRNAs measured by qPCR

Low concentration arsenic exposure increased the expressions of Gclc, Gclm, Keap1 and Bach1 mRNAs, whereas high concentrations reduced their expressions. No significant changes were observed for the expression of Nfe2l2 mRNA following sub-chronic arsenic exposure. * P<0.05; ** P<0.01