Screening differential circular RNA expression profiles reveals the regulatory role of circMARS in anti‐tuberculosis drug‐induced liver injury

Abstract Tuberculosis (TB) treatment is plagued by liver damage, which often leads to treatment interruptions. Circular RNAs (circRNAs) are a special class of non‐coding RNAs abundant in body fluids with important biological functions. However, the role of circRNA in anti‐tuberculosis drug‐induced liver injury (ADLI) is unclear. We explored ADLI‐specific circRNAs in TB patients using circRNA microarrays and verified circMARS in a cohort of 300 individuals. In addition to the value assessment of circMARS in patients using a receiver operating characteristic (ROC) curve, cell experiments were also performed under the guidance of bioinformatics analyses. In particular, we found that circMARS acts as a miRNA sponge by binding to miRNAs. Compared with the blank group, the expressions of circMARS, KMT2C gene, and EGFR protein in the ADLI group were increased, while miR‐6808‐5p, miR‐6874‐3p, and miR‐3157‐5p were decreased. Furthermore, when si‐circMARS was used in the ADLI groups, circMARS demotion manifested the opposite results. Subsequently, a self‐controlled cohort of 35 participants was used to verify the circMARS–miR‐6808‐5p/‐6874‐3p/‐3157‐5p–KMT2C–EGFR function axis. Therefore, circMARS may participate in the compensatory repair mechanism of ADLI through the function axis, and may be a potential biomarker for ADLI diagnosis in TB patients.

drug-induced liver injury (ADLI) hamper the achievement of the treatment goals. Thus, it is of great significance to study the pathogenesis of ADLI. Previous studies have reported that methylation, acetylation, and non-coding RNAs in epigenesis are closely related to ADLI. However, further studies on the role of circRNAs in ADLI are needed. [2][3][4][5] Genetic alterations are the key to revealing the mechanism of ADLI. Circular RNA (circRNA) is a class of novel RNAs commonly transcribed from genome. CircRNA differs from linear RNA, as it has a covalently closed loop structure without the 5′-cap and 3′poly-A tail. 6,7 CircRNA possesses characteristics of universality, stability, specificity, and conservatism; thus, many circRNAs have been found and reported as excellent biomarkers in the past few years. 8,9 CircRNA has been reported to have good potential as a biomarker for malignancies, cardiovascular diseases and active TB. 10 Recent studies have demonstrated that circRNAs have miRNA sponge effects and regulate gene expression by microRNA response elements. 13,14 This study aims to investigate the circRNAs abnormally expressed in TB and to explore their function in ADLI. A two-phase screening/validation project was performed in patients on anti-TB treatment. In the screening phase, the circRNA expression profiles in the serum were assessed using a human circRNA microarray, which was designed to simultaneously detect thousands of circRNAs, in 16 patients with ADLI and 16 non-ADLI patients matched for age, gender, and treatment. In the validation phase, the circMARS in the serum was validated by quantitative real-time polymerase chain reaction (qRT-PCR) in a cohort of 300 patients.
The involvement of circRNAs in liver damage related pathways via interactions with miRNAs was then investigated by multiple bioinformatics approaches. The findings of the bioinformatics analyses were verified in cytology experiments and a self-controlled cohort of 35 patients.

| Subjects design
This study enrolled patients diagnosed with TB at Tangshan Tuberculosis Hospital (China) from July 2015 to July 2018. All patients were put on the same TB therapeutic regimen (daily 2S(E) HRZ/4HR: S, streptomycin; E, ethambutol; H, isoniazid, R, rifampicin; Z, pyrazinamide; dose increased for 2 months and then consolidated for 4 months). ADLI was defined according to the Danan Criteria promulgated in 1990. The inclusion criteria for the ADLI group were liver injury following 6 months of anti-TB drug therapy, while for the non-ADLI group, it was the absence of liver injury following 6 months of anti-TB drug therapy. The patient exclusion criteria were patients diagnosis of other liver diseases, use of other drugs that cause abnormal liver function, and abnormalities in liver structure or function before anti-TB treatment. This study was approved by the Ethics Committee of North China University of Science and Technology. All of the selected patients provided informed consent, and the study protocol conformed to the ethical guidelines of the Declaration of Helsinki (1975).
This study was divided into three phases. Sixteen ADLI patients were selected and matched with 16 non-ADLI patients for age and gender. Total RNAs were then isolated from their peripheral blood for microarray expression profiling. Furthermore, the expression profiles of circRNAs were also detected in the ADLI group. Then, the circRNAs were verified in the second phase, and its diagnos-

| RNA isolation and circRNA microarray expression profiling
Fresh peripheral venous blood samples were centrifuged, and the serum was collected from the upper layer. Subsequently, total F I G U R E 1 Study flow chart. First, the difference of circRNA expression profiles between anti-tuberculosis drug-induced liver injury (ADLI) and non-ADLI patients were explored. Then, a differentially expressed circRNA was selected for verification in a cohort of 300 patients. Finally, the function of this circRNA was verified in a self-controlled cohort of 35 patients based on the results of the experimental study RNA was extracted from the serum and the cell experiments using The circRNA array data were analyzed using the GeneSpring software V13.0 (Agilent). The data were Log2 transformed and median centred by circRNAs using CLUSTER 3.0 software to determine the differentially expressed circRNAs. The differential expression of the circRNAs between both groups was evaluated by assessing the change in the two-fold change (FC > 2) and a T-test p value of 0.05.
The p value was corrected by Benjamini-Hochberg method.  Table S1. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used to normalize the relative expression levels of circRNA and mRNA, and small nuclear U6 was used to normalize the miRNA expression levels. The relative expression levels of the RNA calculation were calculated using the 2 −ΔΔCt method. Furthermore, the PCR products were cut using a gel electrophoresis experiment and were sent to SinoGenoMax (Beijing, China) Co., Ltd., for the sanger sequencing.

| Inferring circRNAs pathways and network
In order to further explore the biological functions of circMARS with a differential expression in ADLI, miRanda 3.0 and TargetScan 7.0 were used to predict the target miRNA of circMARS. Moreover, Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis and Gene Oncology (GO) enrichment analysis of these miRNAs were performed with mirPath v3.0.

| ALT, AST activity and EGFR protein assays
The ALT and AST in the cell supernatant were detected by the Reitman-Frankel method kit (Nanjing Jiancheng Bio, Nanjing, China). The epidermal growth factor receptor (EGFR) concentration in serum was measured by enzyme linked immunosorbent assay (Abcam, Cambridge, UK) according to the manufacturer's instructions.

| Identification of circRNA expression profile in ADLI
In the first phase of the study, circRNA expression profile was de-  (Table S2). All patients were newly diagosed with TB with no previous history of smoking, drinking or other chronic diseases.
The intersection results of the two microarrays showed 113 co-differentially expressed circRNAs, which included seven that were upregulated and 106 that were downregulated (Figure 2b and Table S3).

| Verification of expression levels and the potential value of circMARS
Microarrays have the advantages of high flux and efficiency. However, they are prone to false positives and low stability. Considering the abundance level and biological analysis of circRNA, we selected circ-MARS (circBase ID: hsa_circ_0027252) for qRT-PCR validation. A total of 300 patients were included in the study, with 150 patients allocated to the ADLI group and 150 patients allocated to the non-ADLI group.
There were no significant differences in clinical characteristics (such as age, gender, BMI, smoking, drinking, and education level) between the two groups, as shown in Table 1. Furthermore, the PCR products show that circMARS was resistant to RNase R and sequenced by the Sanger method. The sequence information was consistent with circMARS reported in the circBase database ( Figure S1). The qRT-PCR results of the serum samples showed that the expression of circMARS in the ADLI group was higher than the non-ADLI group, and the difference was statistically significant (Figure 3a). These results suggest that circMARS has a good diagnostic value in ADLI and offers a basis for further studying ADLI-related circRNA.
Results revealed that miRNAs mimics markedly reduced the luciferase activity of the circMARS reporter ( Figure S4). In addition, the CCK8 assay was used to explore the effects of anti-TB drugs and circMARS on the survival rate of hepatocytes. As can be seen from   The expression level of circMARS in the drug-treated groups was much higher than in the blank group (p < 0.05), which was consistent with the results of the circRNA microarray profile (Figure 5c).
As expected, the expression of miR-3157-5p, miR-6808-5p, and miR-6874-3p in the drug-treated groups was decreased compared with the blank group (p < 0.05). Meanwhile, the relative expression of miRNAs in the drug-treated + si-circMARS groups was significantly higher than the drug-treated only groups (p < 0.05; As can be seen from Figure 5f, the EGFR levels in the supernatant of the drug-treated groups were significantly higher than the blank group. However, EGFR was suppressed in the siRNA transfected groups (p < 0.05).

| Verification of circMARS-miRNAs-KMT2C-EGFR function axis in ADLI patients
This study hypothesized that circMARS-miR-3157-5p/-6808-5p/-6874-3p-KMT2C-EGFR function axis is involved in the liver compensatory repair mechanism in ADLI. A total of 35 patients were selected to act as a self-controlled cohort to test this hypothesis. All 35 patients were studied before and during ADLI (the basic information of the patients can be found in Table S5). The relative expression of circMARS and KMT2C mRNA in the serum of patients with ADLI was significantly higher than before ADLI. The relative expression of miR-3157-5p, miR-6808-5p, and miR-6874-3p was significantly lower in the patients' serum during ADLI than before the occurrence of ADLI ( Figure 5). EGFR is involved in the occurrence and progression

| DISCUSS ION
Drug-induced liver injury is a significant obstacle in the treatment of

TB. Previous studies have investigated non-specific factors causing
ADLI, such as interleukin (IL) and tumour necrosis factor (TNF). 2,[17][18][19] However, the underlying mechanism of drug-induced liver injury Accumulating evidence shows that circRNA has been implicated in various biological processes through the regulation of mRNA expression. [26][27][28] CircRNA adsorbs miRNAs, and thus, blocks their inhibitory effect on the target mRNAs. CircRNA-related functional axis has also been confirmed in a variety of diseases.
Guan et al. 29 found that hsa_circ_0016788 regulates hepatocellular carcinoma tumorigenesis through miR-486/CDK4 pathway.  and other liver diseases and the economic cost of exploitation as a diagnostic biomarker should also be evaluated. 34 In conclusion, circMARS may participate in the compensatory repair of anti-TB drug-induced liver injury through the circMARS-miR-6808-5p/-6874-3p/-3157-5p-KMT2C-EGFR function axis.
CircMARS also shows promising potential as a biomarker for ADLI diagnosis in TB patients.

ACK N OWLED G EM ENTS
This research was supported by grants from National Natural Science Foundation of China for Fumin Feng, grant number 81670525.
The funding agencies had no role in the study design, data collection, analysis, or preparation of the manuscript. Thanks are due to Chunyan Meng and Jinfeng Li for assistance with experiments and valuable discussion.

CO N FLI C T S O F I NTE R E S T
The authors confirm that there are no conflicts of interest.