MiR-92a and miR-486 are potential diagnostic biomarkers for mercury poisoning and jointly sustain NF-κB activity in mercury toxicity

Occupational and environmental exposure to mercury is a public health concern worldwide. Although the altered epigenetic regulatory features, such as microRNA, have been associated with mercury exposure, the underlying molecular mechanism is not well illuminated. This study aimed to confirm that hsa-miR-92a and hsa-miR-486 are novel diagnostic biomarkers of occupational mercury poisoning, and to explore the underlying mechanism of miR-92a and miR-486 in mercury toxicity. RT-qPCR assays and receiver operating characteristics curve analyses were conducted to confirm the diagnostic value of miR-92a and miR-486 as biomarkers of occupational mercury poisoning. Dual-luciferase assay was applied to confirm the target gene of miR-92a and miR-486 in vitro. Then, we established an in-vitro model where miR-92a and miR-486 were overexpressed or knocked down in HEK-293 and HUVEC cells. RT-qPCR and western blotting were used to analyze gene and protein expression levels. Cell apoptosis was determined by flow cytometry. Results show that miR-92a and miR-486 expression levels were up-regulated in workers exposed to occupational mercury. Upregulation of miR-92a and miR-486 may play a crucial role in mercury toxicity by jointly activating the NF-κB signaling pathway via targeting KLF4 and Cezanne, respectively.

RT-qPCR analysis. cDNA was synthesized using TaqMan miRNA-specific stem-loop primers and the TaqMan ® miRNA Reverse Transcription Kit (Applied Biosystems, USA), and amplified by PCR. SYBR-green reverse transcription PCR (RT-qPCR) was performed to detect Krüppel-like factor 4 (KLF4), Cezanne, nuclear factor-κB (NF-κB), cyclooxygenase-2 (COX-2) and GAPDH mRNA levels using the SYBR Green PCR Kit (Applied Biosystems, USA). Total RNA was prepared from the cells 24 h after Hg treatment and transfection, followed by RT-qPCR analysis. The sequences and amplification efficiency of each primer can be found in the Supplementary Tables S1-S3.
Western blot analysis. Antibodies against KLF4, Cezanne, NF-κB, COX-2, GAPDH, and secondary antibodies were obtained from Abcam (USA). Primary antibodies and secondary antibodies used for Western blot analyses were diluted 1:1,000 in 5% defatted milk as a working concentration. Western blot analysis was conducted according to the guidelines and instructions of Abcam. GAPDH was used as loading control. Total protein was prepared from the cells 48 h after Hg treatment and transfection, followed by Western blot analysis. 20 μg protein of each sample was loaded onto the SDS-polyacrylamide gels (BIO-RAD). Polyvinylidene-fluoride membrane (0.45 μm) were purchased from and Millipore. and HUVEC cells were co-transfected with pRL-TK vectors, pGL3 vector control, miR-92a, miR-486 mimics or mimic-negative controls, and miR-92a, miR-486 inhibitors or inhibitor negative controls, using lipofectamine ® 2000 (Invitrogen, USA). Firefly luciferase activity was detected using dual-luciferase assays (Promega), which were normalized to a Renilla luciferase expression control. Statistical analysis. All statistical analyses were carried out using SPSS 21.0 software. P-values were determined by ANOV for continuous variables and by chi-square test for categorical variables. LSD were used for multiple comparison in ANOV. Urine mercury underwent a log10-transformation for statistical analysis. The relative expression level of each miRNA was individually calculated by 2 −ΔΔCt (ΔΔCt = (Ct miRNA 1 − Ct control ) − (Ct miRNA2 − Ct control )). The relative quantification value then underwent a log2-transformation to compare the expression levels of the candidate miRNAs between groups. Receiver operating characteristics (ROC) curve analyses and the area under the curve (AUC) analyses were used to evaluate the diagnostic value of miR-92a and miR-486. One-way ANOVA analysis was used to analyze the expression levels of miRNAs, gene and protein levels. In all cases, p < 0.05 was considered statistically significant.

Results
General characteristics of study subjects. As shown in Table 1, the distribution of age, sex, smoking, drinking and working years were matched between A) mercury poisoning group, B) mercury absorbing group and C) control group (all p > 0.05), whereas the urine mercury levels were all significantly different between three groups (p < 0.001).
MiR-92a-3p and miR-486-5p are up-regulated in mercury-exposed workers. As shown in Fig. 1A, expression of miR-92a and miR-486 was significantly up-regulated in workers with mercury poisoning and mercury absorbing workers compared with the control group (P < 0.001). Next, we used receiver operating characteristics (ROC) curve analysis, and calculated the area under the curve (AUC), to evaluate the correlation of miR-92a and miR-486 levels with mercury poisoning and mercury absorption (Fig. 1B). The AUC of miR-92a and miR-486 for workers with mercury poisoning were 0.912 and 0.908, respectively. The AUC of miR-92a and miR-486 for mercury absorbing workers were 0.717 and 0.721, respectively. The diagnostic value of miR-92a is higher than miR-486 for both mercury poisoning and mercury absorption. Subsequently, we sampled urine and blood from the same workers for the next two years. We determined urine Hg, and the expression levels of miR-92a and miR-486 for the group with mercury poisoning and the mercury-absorbing group, as shown in Fig. 1C and D. Scatter diagrams of urine Hg, miR-92a and miR-486 were drew and shown in Fig. 1E and Fig. 1F. Spearman correlation was used to calculate the R 2 and p value of coefficient. The R 2 of correlation between urine Hg and miR-92a was 0.001 and 0.041 for mercury absorbing and poisoning group (P = 0.91 and 0.38). The R 2 of correlation between urine Hg and miR-486 was 0.010 and 0.008 for mercury absorbing and poisoning group (P = 0.67 and 0.71). The R 2 of correlation between urine miR-92a and miR-486 was 0.767 and 0.446 for mercury absorbing and poisoning group (P < 0.001). The relative expression levels of between miR-92a, miR-486 and urine Hg were not statistically correlated.
MiR-92a and miR-486 are up-regulated in Hg-treated (HgT) cells. In order to investigate whether mercury could induce the change of miR-92a and miR-486 expression levels in vitro, HEK-293 and HUVEC cells were treated with mercuric chloride at different concentrations for 24 h. Exposure doses were determined by MTT assays. As shown in Fig. 2A, the survival rates of HEK-293 and HUVEC cells exposed to 10 μM mercuric chloride were 54.28% and 43.03%, respectively. Subsequent RT-qPCR analysis revealed that miR-92a and miR-486 were highly up-regulated in HEK-293 and HUVEC cells (Fig. 2B). Flow cytometry analysis indicated that apoptosis rates of HEK-293 and HUVEC cells were significantly increased after treatment with 5 μM or 10 μM mercuric chloride, compared to the control group ( Fig. 2C and D, respectively).   Table 1. General characteristics of study subjects Ages and working years were shown are mean ± SD; urine mercury was shown as median (P 25~P75 ).
MiR-92a-3p directly targets KLF4 and sustains NF-κB activity. MicroRNA target prediction (TargetScan, miRANDA) showed that KLF4 is the predicted target of miR-92a-3p (Fig. 3A). To explore whether miR-92a definitely targets KLF4, HEK-293 and HUVEC cells were co-transfected with a Renilla luciferase reporter vector containing the wild-type KLF4-3′-UTR (WT) or mutant KLF4-3′-UTR, together with 10 nmol/L miR-92a or a miRNA control. A dual-luciferase reporter gene assay showed that luciferase activity in KLF4 3′-UTR (WT) transfected cells decreased in presence of miR-92a, whereas an inhibitory effect of miR-92a on the luciferase expression in cells containing mutant KLF4-3′-UTR (mut) was not detected ( Fig. 3B and C). Thus, our results demonstrate that KLF4 is a definite target of miR-92a in vitro. Next, miR-92a was overexpressed in HEK-293 and HUVEC cells to investigate the effects of miR-92a on KLF4 expression, and on its downstream signal molecules (NF-κB and COX-2). RT-qPCR confirmed a decreased level of KLF4 mRNA in miR-92a-transfected cells compared with the miRNA control-transfected HUVEC and HEK-293 cells (Fig. 3D). Western blot analysis revealed that the protein levels of KLF4 decreased, whereas phospho-NF-κB p65 and COX-2 increased in miR-92a-transfected cells compared with the control cells (Fig. 3E). Results of a cell apoptosis assay by FCM showed that apoptotic rates increased in miR-92a-transfected cells compared with the controls (Fig. 3F for HEK-293 and Fig. 3G for HUVEC).
MiR-486-5p directly targets Cezanne and activates NF-κB signaling. Based on bioinformatic microRNA target analysis (TargetScan, miRANDA), Cezanne is the predicted target for miR-486-5p (Fig. 4A). HEK-293 and HUVEC cells were co-transfected with the wild-type Cezanne-3′-UTR (WT) or mutant Cezanne-3′-UTR (mut) together with 10 nmol/L miR-486 or miRNA control to prove that miR-486 directly targets Cezanne. The dual-luciferase reporter gene assay showed luciferase activity in cells transfected with Cezanne 3′-UTR (WT) decreased in presence of miR-486, whereas an inhibitory effect of miR-486 on the luciferase expression in cells with mutant Cezanne -3′-UTR (mut) was not detected (Fig. 4B and C). Thus, our results demonstrate that Cezanne is a definite target for miR-486 in vitro. Next, miR-486 was overexpressed in Hg-T (10 μM) HEK-293 and HUVEC cells to investigate the effects of miR-486 on Cezanne expression and its downstream signal molecules (NF-κB and COX-2). RT-qPCR was performed and confirmed the decreased level of Cezanne mRNA in miR-486-transfected cells compared with the miRNA control-transfected HUVEC and HEK-293 cells (Fig. 4D). Western blot analysis revealed that the protein level of Cezanne decreased, whereas phospho-NF-κB p65 and COX-2 increased in miR-486-transfected cells compared with the control cells (Fig. 4E). Results of the cell apoptosis assay by FCM showed that early apoptotic rates increased in miR-486-transfected cells compared with the controls (Fig. 4F for HEK-293 and Fig. 4G for HUVEC).

MiR-92a and miR-486 induce apoptosis by jointly activating NF-κB and COX-2.
In the previous experiments, KLF4 and Cezanne were shown to be the definite targets of miR-92a and miR-486, respectively. Interestingly, numerous studies have reported that KLF4 and Cezanne are both negative regulators of NF-κB, and that inhibition of miR-92a and miR-486 could reduce the expression of the active form of NF-κB, phospho-p65 13 , an important promoter of inflammation 14 . RT-qPCR and Western blot analysis were conducted to prove that miR-92a and miR-486 jointly activate NF-κB, and that inhibition of miR-92a and miR-486 could reduce the expression of NF-κB in Hg-T (10 μM) cells. Hg-T HEK-293 and HUVEC cells were transfected with 10 nmol/L miR-92a and miR-486 or 20 nmol/L anti-92a and anti-486. We observed that mRNA and protein levels of KLF4 and Cezanne in miR-92a and miR-486 transfected cells were decreased compared to levels in Hg-T cells, but increased in anti-92a and anti-486 transfected cells (Fig. 5A and B). In addition, phospho-NF-κB p65 and COX-2 mRNA and protein levels were markedly increased in miR-92a and miR-486 transfected cells compared to levels in Hg-T cells; however, they were decreased in anti-92a and anti-486 transfected cells (Fig. 5C-E). Cell apoptosis analysis showed that apoptotic rates increased in miR-92a and miR-486 transfected cells compared with Hg-T cells. In contrast, anti-92a and anti-486 could reduce the cell apoptotic rates compared with miR-92a and miR-486 transfected cells (Fig. 5F for HEK-293 and Fig. 5G for HUVEC). Fig. 6A and B, COX-2 mRNA and protein levels were markedly reduced in siRNA1 and siRNA2-COX-2 transfected HEK-293 and HUVEC cells compared to levels in Hg-T cells. In Fig. 6C, siRNA1 and siRNA2 of COX-2 could reduce the Hg-induced apoptotic rates of HEK-293 and HUVEC cells compared with the controls.

Discussion
The associations between the aberration of miRNAs profiles and heavy metals exposure have been confirmed by a number of studies 15,16 . Our population-based study confirmed that miR-92a and miR-486 expression levels were upregulated in workers exposed to occupational mercury. Subsequent experimental research revealed that exposure to mercury induced an increase in miR-92a and miR-486 expression, which led to increased NF-κB and COX-2 activation and apoptosis of cells. NF-κB signaling plays a critical role in mediating the expression of numerous proinflammatory gene products, including COX-2 and inducible nitric oxide synthase (iNOS), which are involved in inflammatory responses 17 . Consistent with our findings, Park et al. 17 reported recently that mercury could induce NF-κB activation, resulting Each bar represents the mean ± SE of three independent experiments. (D) Real-time PCR analysis of KLF4 expression in miR-92a or miR-con transduced cells. Each bar represents the mean ± SE of KLF4/GAPDH of three independent experiments. (D) Western blot analysis of KLF4, phospho-NF-κB p65 and COX-2 protein levels in miR-con or miR-92a transduced cells. GAPDH served as the loading control. Apoptosis rates of miRcon or miR-92a transduced HEK-293 (F) and HUVEC (G) cells were determined by flow cytometry. *p < 0.05, compared with control.
in increased expression of COX-2. However, previous research did not thoroughly illuminate the relationship between mercury exposure and NF-κB activation. Here, we demonstrate that miR-92a and miR-486 may be the bridge between mercury exposure and NF-κB activation.
In this study, miRNA target prediction and experimental results identified KLF4 and Cezanne as direct targets for miR-92a and miR-486, respectively, which was consistent with findings reported previously 18,19 . Interestingly, KLF4 and Cezanne have both been reported to be important negative regulators of NF-κB signaling, and they repress phospho-NF-κB-p65 (the active form of NF-κB) through different mechanisms. KLF4, one of the zinc-finger transcription factors that mediate inflammation, can interact with phospho-NF-κB-p65 and promote COX-2 activity 20 . However, Cezanne, a newly identified member of the A20 family of deubiquitinases, suppresses NF-κB activity by deconjugating K63-polyubiquitin chains from RIP1 and TRAF6 19 . Consistently, our research demonstrated that overexpression of miR-92a and miR-486 can jointly increase NF-κB activity and COX-2 expression by targeting KLF4 and Cezanne, respectively, whereas inhibition of miR-92a and miR-486 showed the opposite effects.
MiR-92a, possibly originated from endothelial cells, has been proven to play a predominant role in re-endothelialization, angiogenesis, and inflammation 21,22 . It is also an important regulator of atherosclerosis development through endothelial activation 23 . Thus, previous studies together with our results suggest that miR-92a may a play crucial role in mercury-driven hypertension, atherosclerosis and other cardiovascular diseases. A recent study showed that miR-486 is a critical mediator involved in NF-κB activation, which contributes to glioma progression 19 . COX-2 was found to be markedly upregulated in brain disease and could induce apoptosis Apoptosis rates of miR-con or miR-486 transduced HEK-293 (F) and HUVEC (G) cells were determined by flow cytometry. *p < 0.05, compared with control. through its downstream product prostaglandin E 2 (PGE 2 ) 24,25 . All of the previous evidence combined with our results was helpful to explain the neurotoxicity resulting from mercury exposure. However, the mechanisms by which miR-92a and miR-486 are upregulated in mercury exposed workers are still unknown and currently under investigation in our laboratory.
Future studies will include follow up on the workers investigated in this study to observe the effects of occupational mercury exposure on the cardiovascular system, and aim to identify the underlying mechanism. In addition, there was no statistical correlation between the relative expression levels of miR-92a, miR-486 and urine Hg over time. The expression pattern of plasma miR-92a and miR-486 of mercury exposed workers over time merits future research. In our study, only one target was selected for validation for each miRNA, thus validation of more targets of miR-92a and miR-486 and exploration of the role of these targets is one of the important directions of future research. Last, males only make up just 13.3% of all the subjects, conclusions of this population study may be only appropriate for female workers exposed to mercury. Future study including a larger proportion of male is needed to prove our results.

Conclusions
The present study provides a novel mechanism of mercury toxicity and may be the first study in which the role of miRNA in mercury toxicity is well illustrated. As shown in Fig. 6D, exposure to mercury could induce the upregulation of miR-92a and miR-486 which results in reductions of KLF4 and Cezanne. In addition, reductions of KLF4 and Cezanne jointly induce NF-κB activation and COX-2 expression, which eventually cause apoptosis. These combined regulatory effects of miR-92a and miR-486 in NF-κB activation may have crucial roles