Dysregulated circRNAs in plasma from active tuberculosis patients

Abstract Endogenous circular RNAs (circRNAs) have been reported in various diseases. However, their role in active TB remains unknown. The study was aimed to determine plasma circRNA expression profile to characterize potential biomarker and improve our understanding of active TB pathogenesis. CircRNA expression profiles were screened by circRNA microarrays in active TB plasma samples. Dysregulated circRNAs were then verified by qRT‐PCR. CircRNA targets were predicted based on analysis of circRNA‐miRNA‐mRNA interaction. GO and KEGG pathway analyses were used to predict the function of circRNA. ROC curve was calculated to evaluate diagnostic value for active TB. A total of 75 circRNAs were significantly dysregulated in active TB plasma. By further validation, hsa_circRNA_103571 exhibited significant decrease in active TB patients and showed potential interaction with active TB‐related miRNAs such as miR‐29a and miR‐16. Bioinformatics analysis revealed that hsa_circRNA_103571 was primarily involved in ras signalling pathway, regulation of actin cytoskeleton, T‐ and B‐cell receptor signalling pathway. ROC curve analysis suggested that hsa_circRNA_103571 had significant value for active TB diagnosis. Circulating circRNA dysregulation may play a role in active TB pathogenesis. Hsa_circRNA_103571 may be served as a potential biomarker for active TB diagnosis, and hsa_circRNA_103571‐miRNA‐mRNA interaction may provide some novel mechanism for active TB.

expression by buffering their repression of mRNA targets 4 or sequestering specific miRNAs, 5 which have drawn increasing interest, especially with the discovery of their tissue-specific, cell type-specific or developmental-stage-specific expression. 6 Compared with traditional linear RNA, circRNAs with covalently closed continuous loops have more resistance to RNase degradation, which allow them to be selectively enriched during sample processing and make them more suitable for using as good candidate of molecular diagnostic biomarkers than other types of RNA. 7 Numerous publications have demonstrated that aberrant expression of circRNAs has been shown to occur in various human diseases. Studies reported the association of hsa_circ_0002062 with pulmonary hypertension, 8 the increased expression of hsa_-circRNA_104871 with rheumatoid arthritis diagnosis 9 and HIF1aassociated circRNA promoting the proliferation of breast cancer cells. 10 These findings indicate that circRNAs may be significant biological molecules to understand disease mechanisms and to identify biomarkers for disease diagnosis and therapy. Recently, the involvement of circRNAs in viral infection has also been explored. 11,12 The data suggest that circRNAs may play critical role in fine-tuning immune response against microbial infection. 13 However, little is known about the role of circRNAs in active TB.
So, microarray screening and bioinformatics were combined to investigate the differential expression profile of circulating circRNAs in active TB patients in the study. Altered changes of circRNAs may lead to the understanding of potential role of circRNAs in active TB pathogenesis and diagnosis.  After washing the slides, the arrays were scanned by an Agilent Scanner G2505C.

| Microarray data analysis
Scanned array images were imported into Agilent Feature Extraction software (version 11.0.1.1) for raw data extraction. Quantile normalization of raw data and subsequent data processing were performed using R software limma package. After quantile normalization of the raw data, low intensity filtering was performed, and the circRNAs that at least 1 out of 6 samples have flags in "P" or "M" ("All Targets Value") were kept for subsequent differential analysis. Differentially expressed circRNAs between case and controls were identified through fold change (FC) filtering and Student's t test. A false discovery rate-(FDR) adjusted P value was calculated. Specifically, only the circRNA that exhibited FC > 1.5 and P value < .05 was selected as differentially expressed ones. Scatter plot was used to assess the variation in circRNA expression, volcano plot was applied to visualize the differential circRNAs between the two groups, and hierarchical clustering was performed to show the distinguishable circRNA expression pattern among the samples.

| qRT-PCR
qRT-PCR was used to confirm circRNA expression and detect potential target miRNA of circRNA. Following RNA extraction from active TB and control plasma, cDNA was synthesized with SuperScript For miRNA analysis, PCR reaction was performed in a total volume of 20 lL, including 2 lL cDNA, 2.5 lL 2 9 SYBR Green, 1 lL primer forward (10 lmol/L) and 1 lL primer reverse (10 lmol/L) (if needed, their sequences were available). The cycling programme consisted of 95°C for 5 minutes, followed by 40 cycles of 95°C for 10 seconds, 60°C for 60 seconds and 72°C for 20 seconds. U6 was used as internal control. All PCR reactions were conducted in triplicate. Relative expression was calculated using the 2 ÀDDCt method. Control reference value for each circRNA and miRNA was obtained from pooled RNA sample from 29 healthy controls.

| Annotation and function prediction for the validated circRNAs
To further elucidate the correlations between circRNA and miRNA, we predicted circRNA/miRNA interaction using miRNA target prediction software from Arraystar, which refers to TargetScan (http:// www.targetscan.org/) and miRanda (http://www.microrna.org/). The Arraystar software was used to search for miRNA response element (MREs) on the qPCR differential circRNAs, which were then used to identify putative miRNAs based on their seed sequence complementarity. MiRNA-targeted mRNAs were predicted by TargetScan and miRanda, GO (Gene Ontology) and KEGG (Kyoto Encyclopedia of F I G U R E 1 Scatter plots and volcano plots to identify differentially expressed circRNAs between active TB samples and controls. A, circRNA Scatter Plot. The values spotted in the X and Y axes represent the normalized signals of samples in the two groups. The green lines represent fold changes. The circRNAs above the upper green line and below the lower green line are those with expression fold changes >1.5 between the two groups. B, circRNA Volcano plot. The vertical lines correspond to 1.5-fold up and down, respectively, and the horizontal line represents a P value of .05. The red point in the plot represents the differentially expressed circRNAs with P < .05. BC: active TB group; CC, control group F I G U R E 2 Hierarchical cluster of differentially expressed circRNAs between active TB samples and controls. The colour scale runs from green (low intensity), through black (medium intensity), to red (strong intensity). BC: active TB group (n = 3), including BC1, BC2 and BC10; CC: healthy control group (n = 3), including CC6, CC7, CC8 samples Genes and Genomes) pathway analyses were conducted to predict hsa_circRNA_103571 function based on analysis of its targeted miRNA-mRNAs.

| ROC curve analysis
A receiver operating characteristic curve (ROC) was plotted; the area under the curve (AUC), sensitivity and specificity were calculated to assess the value of dysregulated circRNA for active TB diagnosis.

| Statistical analysis
The groups were compared to evaluate the statistical significance using Student's t test or Chi-square test, as appropriate. Correlation between circRNAs and their predicted miRNA targets for individual donor samples were evaluated. P < .05 was considered statistically significant.

| Identification of dysregulated circRNAs expression profile in active TB patients
To determine the differential circRNA expression associated with TB, three patients with active TB and three healthy controls (male/female = 2/1) were selected as the circRNA array cohort. Microarray analyses are affected by FDR as a consequence of multiple testing.
Therefore, adjustment of P values for multiple testing has to be done. However, FDR adjusted P values for all circRNAs are more than .05 in the study, which means that the initial array study was underpowered to detect significant differences and more patient samples should be included for future studies of multiple circRNAs.
So we presented P value without adjustment in the study. A total of 75 differentially expressed circRNAs were identified based on FC > 1.5 ( Figure 1A) and P < .05 ( Figure 1B), and among them, 47 F I G U R E 3 qRT-PCR analysis of circRNAs and miRNAs A, Validation of selected dysregulated circRNAs. Expression levels of hsa_circRNA_091692, hsa_circRNA_102296 and hsa_circRNA_029965 were higher, while hsa_circRNA_103571 was lower in active TB (n = 32) compared with controls (n = 29). b-actin was used as internal control. B, Detection of potential target miRNAs of hsa_circRNA_103571. Plasma hsa-miR-16 and hsa-miR-29a were increased in active TB (n = 32) compared with controls (n = 29). C, hsa_circRNA_103571 expression was negatively correlated with level of hsa-miR-29a (r = À.63, P < .05) and hsa-miR-16 (r = À.65, P < .05) in the patient group, respectively. U6 was used as internal control. The data were analysed using the 2 ÀDDCt method and presented as relative expression levels from three independent experiments. Student's t test was used for statistical analysis. *P < .05 vs controls circRNAs were up-regulated and 28 were down-regulated in active TB patients, respectively. A hierarchical cluster was created to group the dysregulated circRNAs among the samples (Figure 2). Among these circRNAs, hsa_circRNA_051239 and hsa_circRNA_406841 were the top increased and the most decreased ones, respectively.
When the filter criteria were set as FC > 2 and P < .05, 16 circRNAs were significantly elevated and only one was obviously reduced in active TB group compared with healthy controls (

| Validation the differential expressed circRNAs
To further assess the microarray expression data, we selected 6 dys- for validation based on the following criteria: raw intensity >100, exonic-related circRNAs, FC > 1.5 and P < .05. A total of 32 active TB patients and 29 healthy controls were enrolled as a validation cohort. qRT-PCR method was used to confirm the differential expression levels from participant's plasma samples. Consistent with the microarray data, hsa_circRNA_091692, hsa_circRNA_102296, hsa_circRNA_029965 and hsa_circRNA_103571 were significantly dysregulated between cases and controls, while the expression of hsa_circRNA_406755 did not display any remarkable difference between the two groups ( Figure 3A). The expression level of hsa_-circRNA_100823 was not detectable in plasma sample from healthy controls.

| Annotation and function prediction for hsa_circRNA_103571
Recent evidences have demonstrated that some circRNAs can perform the function of "miRNA sponge" by efficiently binding and inhibiting miRNA transcription, which would further influence downstream mRNA expression and finally involved in various diseases. 14 In the study, to investigate the potential interactions between cir-cRNAs and its possible targeted miRNAs, we searched for MREs targeted by qRT-PCR validated differentially expressed circRNAs. As present in Table 1, dominant MREs were targeted by the dysregulated circRNAs. Some of these miRNAs have been reported to be associated with active TB, such as miR-29a and miR-16, which were up-regulated during active TB infection. It was worth noting that these miRNAs above mentioned were targeted by hsa_-circRNA_103571, which was decreased in active TB group vs healthy controls in our study. We further strengthened the assumption by qRT-PCR. As shown in Figure 3B (Table S3).

| Receiver operating characteristic analysis
To assess the potential value of confirmed hsa_circRNA_103571 for active TB diagnosis, we further performed ROC curve analysis. From this analysis, we found that ROC curve of hsa_circRNA_103571 showed a significant distinguishing efficiency with an AUC value of 0.838 (95% CI: 0.734-0.941, P < .001) ( Figure 5), which indicated that hsa_circRNA_103571 could separate patients with active TB from healthy controls.
YI ET AL.

| 4081
and hsa_circ_0005836 may be served as a novel potential biomarker for TB diagnosis. 15 Another study has demonstrated that cir-cRNAs are involved in macrophage response to Mtb infection. 16 These findings suggest that circRNAs are worth studying for better In the study, we characterized the profile of differentially expressed circulating circRNAs in active TB patients vs healthy controls, and we found that the number of up-regulated circRNAs was more than that of down-regulated ones. Noticeably, when filter criteria were set as FC > 2 and P < .05, 17 circRNAs were identified with differential expression between samples from subjects with or without active TB, of which only one was significantly reduced in active TB group. This indicated that circulating circRNAs showed a tendency to be increased upon TB infection. Functions of circRNAs remain largely unclear. CircRNAs may be generated from exonic or intronic sequence and function as modifiers of parental gene expression. 17,18 In the current study, we sequenced four circRNAs of interest following a microarray screening.
Among them, hsa_circ_091692, hsa_circ_102296 and hsa_circ_029965 were up-regulated in active TB plasma. Hsa_circ_091692 is encoded by the CD99L2 (CD99-Like 2) gene, the protein product expressed on leucocytes and play an important role in various acute and chronic inflammatory diseases. 19 Hsa_circ_102296 aligns with ANKRD12 (Ankyrin repeat domain 12), which encodes a 224 kD nuclear protein associated with colorectal cancer. 20 Hsa_circ_029965 is derived from PDS5B, and its encoded protein is related to chromosome segregation. 21 On the other hand, hsa_circ_103571 was down-regulated in active TB plasma. Hsa_circ_103571 aligns with the gene LRCH3, which encodes a protein that may be involved in resistance to bacterial infection. 22 However, roles of these circRNA-related protein-coding genes in active TB are largely unknown.
In addition to action as modifiers of parental gene expression, other intriguing possibility is that circRNAs might regulate miRNA function as miRNA sponges. Evidences are arising that circRNAs associate with related miRNAs and the circRNA-miRNA-mRNA axes are involved in many disease pathways. For example, it has been reported that hsa_circ_0010729 can regulate vascular endothelial cell proliferation and apoptosis via targeting miR-186/HIF-1a axis. 23 So, we assumed the same mechanism for the dysregulated circRNA regulation in the development of active TB. We were interested in hsa_circ_103571, which was decreased and had good diagnostic accuracy for active TB in our study. Bioinformatics result showed that it was matched with hsa-miR-29a and hsa-miR-16, which have been reported to be up-regulated in active TB. 24,25 Moreover, we found that hsa-miR-29a was also matched with hsa_-circRNA_104674, which was down-regulated in our study. Other miRNA, such as has-miR-33, which were matched with down-regulated hsa_circRNA_100237 in our study, has also been reported to be increased in Mtb infection. 26 These findings suggested that specific circRNAs, such as hsa_circ_103571, might be involved in TB pathogenesis by binding and inhibiting related miRNA transcription.
The differential expressed hsa_circ_103571 found in active TB has not been reported in the other diseases ever since, which indicates the specificity and efficiency of this circRNA as a molecular biomarker.
By bioinformatics analysis of circRNA-miRNA-mRNA interaction, we predicted the function of hsa_circ_103571 in active TB. Results from GO process and KEGG pathway could help to enrich and identify important mRNAs. In the top dysregulated GO processes of differentially expressed mRNAs, the biological process of autophagy, fatty acid biosynthetic process and regulation of actin cytoskeleton organization were identified. As autophagy plays an important role in the development of active TB, 27 it is possible to speculate that mRNAs might be jointly involved in this process, and the predicted mRNAs in circRNA-miRNA-mRNA network might be related to hsa_circ_103571 function.
KEGG enrichment analysis also points to the effective signalling pathways, such as regulation of actin cytoskeleton, which has been reported to be a key mediator of mycobacterium infection. 28 In summary, our study provided a profile of circulating circRNAs in active TB and healthy controls. The alteration of circulating cir-cRNAs expression indicated that these dysregulated circRNAs might be involved in the onset and development of active TB. Specific cir-cRNAs may regulated TB-related miRNAs, and thus might play a role in the pathogenesis of active TB and might represent novel biomarkers for this disease. However, available datasets are limited and these circRNAs signatures should be further confirmed in future studies.

ACKNOWLEDG EMENTS
The present study was supported by the Natural Science Foundation