Peripheral blood samples (2 ml) were collected from 30 patients at the People's Hospital of Guangxi Zhuang Autonomous Region between January 2016 and December 2017. Combined with the clinical index and the manifestation of the patients, the diagnosis of IMN was confirmed by renal biopsy and immunofluorescence. Inclusion criteria for patients with IMN were as follows: Proteinuria, with some patients having mild microscopic hematuria or renal insufficiency; diffuse thickening of the glomerular capillary wall and deposition of immune complexes in the subepithelial spaces, as observed by light microscopy; immunofluorescence results showing IgG4 deposited along the capillary loops, which could be combined with deposits of complement component 3 (C3); electron-dense deposits on the epithelial side or glomerular basement membrane, and foot process fusion by electron microscopy; and when the blood samples were collected, all the patients with IMN included in the disease group had not received glucocorticoid therapy or other immunosuppressive agents. The exclusion criteria for the IMN group were as follows: Patients with autoimmune diseases, including systemic lupus erythematosus, rheumatoid arthritis, multiple sclerosis, idiopathic thrombocytopenia or type 1 diabetes; patients with acute or chronic infectious diseases, including hepatitis virus, AIDS, inflammatory enteritis, parasitic infection and other acute or chronic infectious diseases; patients with cancer, a kidney transplant or heavy metal poisoning; and patients diagnosed with IMN who had previously received treated.

In total, 30 specimens were obtained from individuals who underwent annual checkups and were confirmed to be healthy between January 2016 and December 2017 at the medical examination center of the People's Hospital of Guangxi Zhuang Autonomous Region. These 30 specimens formed a control (NC) group. The following criteria were adopted to select the samples in the NC group: The age and gender were consistent with the patient group; the subjects had no other comorbidities; it was known that clinical test indexes, including renal function, were normal; and there were no clinical manifestations related to other diseases.

Blood samples were stored at −80°C immediately until further analysis. The protocol for this research was reviewed and approved by the Ethics Committee of the People's Hospital of Guangxi Zhuang Autonomous Region. All participants of the present study signed an informed consent form.

Total RNA in the peripheral blood was extracted using TRIzol reagent (Invitrogen; Thermo Fisher Scientific, Inc.), according to the manufacturer's protocol. The purity and concentration of the RNA were determined using a NanoDrop ND-1000 (Nanodrop Technologies; Thermo Fisher Scientific, Inc.), and the integrity of the RNA was determined using denaturing agarose gel electrophoresis. The RNA was stored at −80°C for the following experiments.

Each microarray contained 6 samples (3 patients with IMN and 3 controls), which were also included in the 30 samples used for PCR validation. Sample labeling and array hybridization were performed according to the manufacturer's protocol (Arraystar, Inc.). Briefly, total RNA was digested with RNase R (Epicentre; Illumina, Inc.) to remove linear RNAs. Then, the enriched circRNAs were amplified and transcribed into fluorescent complementary (c)RNA using a random priming method (Arraystar Super RNA Labeling kit; Arraystar, Inc.). The labeled cRNAs were hybridized onto the Arraystar Human circRNA Array V2 (8×15K; Arraystar, Inc.). After the slides were washed, the arrays were scanned using the Agilent Scanner G2505C (Agilent Technologies, Inc.). Scanned images of the raw data were imported into Agilent Feature Extraction software (version 11.0.1.1; Agilent Technologies, Inc.). Quantitative normalization of the raw data and subsequent data processing were performed using the R software package (version 3.1.2; http://cran.r-project.org/). The statistical significance of the differences were estimated using unpaired t-tests. CircRNAs having fold change (FC)>2.0 and P<0.01 were selected as significantly differentially expressed and were displayed by Volcano Plot filtering. miRNA targets of circRNAs and circRNA/miRNA interactions were predicted by the Arraystar Human circRNA Array V2 computer program based on TargetScan 7.1 (21) and miRanda (http://mirdb.org) (22).

To further elucidate associations between circRNAs and miRNA, circRNA/miRNA interactions were predicted using Arraystar's miRNA target prediction software (Arraystar, Inc.) based on TargetScan 7.1 and mirdb 5.0. All differentially expressed circRNAs were annotated in detail with the circRNA/miRNA interaction information. GO analysis was (23,24) performed to explore the functional roles of target genes in terms of ‘biological processes’ (BPs), ‘cellular components’ and ‘molecular functions’ (MFs). Biological pathways were defined using KEGG (25,26) and identified using Database for Annotation, Visualization and Integrated Discovery (27,28)

RT-qPCR analysis was performed to verify the results of the microarray analysis for five differentially expressed circRNAs. For RT reactions, 2 µg RNA, 1 µl 0.5 µg/µl random primers (N9) and 1.6 µl dNTP Mix (2.5 mM; HyTest, Ltd and H₂O to a volume of 14.5 µl was incubated at 65°C for 5 min and then placed on ice for 2 min. After this, 4 µl 5X First-Strand Buffer, 1 µl 0.1 M DTT, 0.3 µl RNase Inhibitor (Epicentre; Illumina, Inc.) and 0.2 µl SuperScript III (Invitrogen; Thermo Fisher Scientific, Inc.) were added and incubated at 37°C for 1 min. A qPCR Master Mix (Arraystar, Inc.) and the ViiA 7 Real-Time PCR system (Applied Biosystems; Thermo Fisher Scientific, Inc.) were used for qPCR, following the manufacturer's instructions. qPCR was performed as follows: 95°C for 10 min followed by 40 cycles of 95°C for 10 sec and 60°C for 60 sec. Divergent primers, instead of the more commonly used convergent primers, were designed for the five selected circRNAs. β-actin was used as the internal control for circRNA normalization and quantification. The relative expression of each circRNA was analyzed by the 2^−ΔΔCq method (29). All primers were synthesized by Invitrogen; Thermo Fisher Scientific, Inc. The results are expressed as the mean ± SEM of three independent experiments. Primer sequences are shown in Table SI.

To explore the potential biological function of the candidate circRNA, a circ_101319-targeted miRNA-mRNA network was conducted using Cytoscape (version 3.6.0) (30), according to the analysis of TargetScan and miRanda. The target miRNAs for circ_101319 were identified and subsequently ranked based on their miRNA support vector regression (mirSVR) scores; the source code for the mirSVR scoring system was provided by Aksomics, Inc. The size of each node represents the different types and numbers of putative targets functionally connected to each circRNA. GO annotations and KEGG pathways were used to analyze the predicted gene functions in the network.

Kidney tissues from patients were obtained using renal puncture. A part of each tissue sample was sliced to 3–5 µm using freezing microtome (−20°C) and then cut into frozen sections. Immunofluorescence staining of IgA, IgG, IgM, C3, complement component 1q (C1q) was performed on the sections. The following fluorescently-labeled rabbit anti-human antibodies (1:40; Gene Tech Biotechnology Co., Ltd.) were used: IgG (cat. no. GF020229), IgA (cat. no. GF020429), IgM (cat. no. GF02032), C3 (cat. no. GF020129), C1q (cat. no. GF025429). Antibodies were added separately to each specimen and incubated in at 37°C for 40 min. Samples were washed three times with PBS for 3 min each time. The sections were observed under a fluorescence microscope (magnification, ×200), images were captured and the results were recorded. The remaining parts of the renal specimens were fixed using glutaraldehyde (2.5%) at 4°C for 2 h and embedded in epoxy resin. Samples were dehydrated using an ethanol series (50, 70, 80, 90 and 95%) and then incubated in propylene oxide (100%) for 45 min at 20°C. Ultra-thin sections (50–70 nm) were prepared after polymerization. The prepared sections were stained with uranyl acetate (3%) for 30 min at 25°C and lead citrate (3%) for 8 min at 25°C. Pathological changes in the renal tissue were observed and photographed using a transmission electron microscope.

The normalized data from the microarray processing were analyzed using the R software package (version 3.1.2, http://cran.r-project.org/) and P-value were calculated using unpaired t-tests. All other statistical data were visualized and analyzed using GraphPad Prism 5.0 (GraphPad Software, Inc.). Microarray and RT-qPCR data are presented as the means ± SEM, and are representative of at least three experiments. Differences between groups were evaluated using a two-tailed Student's t-test. P<0.05 was considered to indicate a statistically significant difference. Analysis of ROC curves was applied, and the area under the curve was calculated with 95% confidence intervals (CIs), which were used to assess the prediction value of the circRNAs in disease.

Between January 2016 and December 2017, >100 patients with renal diseases were screened. The results of renal biopsies and immunofluorescence analysis in these patients were observed. Combined with their clinical manifestation, 30 patients were diagnosed with IMN and were selected for the present study. The baseline characteristics are shown in Table I. The average age of the patients who underwent renal biopsy was 53±11 years, and 63% of the patients were male. Among the laboratory indexes listed, the level of proteinuria was 7.6±3.0 g/24 h, and urine and serum albumin was 23±3.8 g/l, which could reflect the nephrotic syndrome. Endogenous creatinine clearance was 81±24 ml/min per 1.73 m², indicating that the kidney function of patients was well preserved. In addition, the mean level of serum cholesterol was 10±3.4 mmol/l. All the electron microscopy images and fluorescence analysis from the patients in disease group were consistent with IMN.

High-throughput microarray analysis of human peripheral blood was used to assess circRNA expression signatures in IMN. A total of 13,617 circRNA targets were identified using the microarray probes in three pairs of clinical specimens. In total, three microarrays were analysed for each group, with each microarray containing pooled RNA from two people. Box plots were used to visualize the distribution of intensities from all normalized data, and it was found that the distribution of circRNAs displayed using log2 ratios was almost equal in the IMN and NC groups (Fig. 1A). FC filtering revealed the differentially expressed circRNAs (Fig. 1B). In addition, the differentially expressed circRNAs with statistically significant differences between the two groups were identified by volcano plot filtering (Fig. 1C). The results of the microarray analysis showed that 955 circRNAs were differentially regulated (FC≥2.0; P<0.05; false discovery rate <0.05), of which 645 circRNAs were upregulated and 310 circRNAs were downregulated. Among these circRNAs, upregulated circRNAs were more prevalent than downregulated circRNAs and were more closely related to disease. The distribution of circRNAs on human chromosomes is described in Fig. 1D. The majority of differentially regulated circRNAs are transcribed from the exonic regions of protein coding genes (Fig. 1E). Certain circRNAs are also transcribed from intronic regions, and some are sense overlapping. Few circRNAs are transcribed from other sources (Fig. 1E). The top 10 upregulated and downregulated circRNAs identified by microarray analysis are listed in Table II. Hierarchical clustering analysis showed that the expression of circRNAs in peripheral blood was significantly different between the NC group and patients with IMN (Fig. 2).

To further determine the biological function and potential regulators of circRNAs, software made by Arraystar to predict the miRNA targets of differentially expressed circRNAs based on bioinformatics analysis by TargetScan and miRanda databases was used. According to the analysis of complementary miRNA matching sequences and the microarray results, circRNA-miRNA interaction networks were constructed using Cytoscape. The interaction maps, including the top 10 upregulated and downregulated circRNAs and their corresponding miRNAs, are shown in Fig. 3. The five microRNA response elements with good mirSVR scores (a lower score indicates a more stable combination of miRNA and mRNA) for the top 10 dysregulated circRNAs are shown in Table II. Additionally, the host genes and genomic locations of these circRNAs are also provided for further functional prediction.

To explore the potential regulatory roles of circRNAs and related molecules in IMN, GO analysis of the predicted targets from the dysregulated circRNAs was performed. It was found that the BPs of the genes corresponding to the top 10 upregulated circRNAs were involved in the ‘positive regulation of the cellular metabolic process’ and ‘positive regulation of cellular process’, ‘positive regulation of the RNA metabolic process’ and ‘system development’. Their MFs were enriched in ‘regulatory region nucleic acid binding’, ‘transcription regulator activity’, ‘RNA polymerase II transcription factor activity’ and ‘protein binding’. Analysis of the downregulated circRNAs showed that the BPs included ‘system development’ and ‘positive regulation of cellular process’, whereas the MFs were consistent with the upregulated genes, with the exception of ‘enzyme binding’ (Fig. 4A and B). Following GO annotation, KEGG pathway analysis was used to identify the intermolecular signaling networks, which was informative for elucidating and interpreting the potential relationships between circRNAs and IMN pathogenesis. It was found that the mitogen-activated protein kinase (MAPK) signaling pathway, the transforming growth factor (TGF)-β signaling pathway and cellular senescence were ranked highly in the upregulated circRNA group, whereas the MAPK, Ras and Rap1 signaling pathways were ranked highly, with high enrichment scores, in the downregulated group (Fig. 4C and D).

After the bioinformatics analysis on the circRNA/miRNA network and determination of the potential relationships between these molecules and the mechanisms underlying IMN, five candidate circRNAs were selected for further validation using RT-qPCR. At this stage, the sample size was increased by collecting the peripheral blood from 30 subjects in each group. The results showed no marked difference between the IMN and NC groups in terms of the expression of circ_033475, circ_102355, circ_101854 or circ_102711, whereas the expression of circ_101319 was significantly higher in the IMN group (Fig. 5A). This trend was consistent among the microarray and RT-qPCR analysis. The expression of circ_101319 in the IMN and NC groups is shown in Fig. 5B. These results suggested that there was a larger degree of variation in the expression of circ_101319 in the IMN group compared with the NC group.

It was found that the expression of circRNAs in peripheral blood was different between the IMN and the NC groups via microarray analysis, and RT-qPCR analysis supported the upregulation of circ_101319 in the IMN group. Due to the characteristics of circRNAs, including resistance to exonucleases and degradation, and stable expression, circRNAs may have an important role to play in molecular diagnostics. ROC curves were used to explore the association between circ_101319 expression and IMN. When circ_101319 expression was applied for the purposes of this analysis, the area under the ROC curve (AUROC) reached 0.89 (95% CI; 0.815–0.971), suggesting that circ_101319 may be used as a predictor for IMN. Moreover, the results showed that the sensitivity and specificity of circ_101319 were 93.33 and 70.00%, respectively (Fig. 6), suggesting that the upregulated expression of circ_101319 was associated with the diagnosis of IMN, which is supported by electron microscopy and immunofluorescence images from renal biopsies (Fig. 7). The glomerular basement membrane (GBM) was thickened and the electron-dense deposits were observed on the epithelial side of the GBM by TEM. Immunofluorescence results showed that IgG4 was granular positive and distributed along the glomerular capillary loops. Negative staining results for IgM and IgA was observed and there was no deposition of complement components C4 or C1q in patients with IMN.

After clinical observation, the target miRNAs for circ_101319 were identified and subsequently ranked based on their mirSVR scores. The five highest ranking candidate miRNA binding targets were used for further analysis of underlying molecular mechanisms through specific base pairing. Combined with previous research concerning circRNA-mediated regulation, it was hypothesized that circ_101319 may also regulate a circRNA-miRNA-mRNA network through a competitive endogenous RNA mechanism (31–33). The potential interactions were predicted using TargetScan and miRanda. Based on these analysis tools, a map of five miRNAs and their predicted target mRNAs was constructed to demonstrate the circRNA-miRNA-mRNA network of circ_101319, which included miR-135a, −5p and −3p, miR-138-5p and miR-338-3p (Fig. 8). The co-expression network of circ_101319 indicated that the miR-135 family exhibited the largest interaction network. To gain further insight into the potential functions of circ_101319 in renal disease, GO annotation and KEGG pathway analysis were applied to the results. It was found that circ_101319 was involved in ‘cellular senescence’, ‘renal cell carcinoma’ and ‘renin secretion’ (Fig. 9).