Elevated hsa‐miR‐590‐3p expression down‐regulates HMGB2 expression and contributes to the severity of IgA nephropathy

Abstract Peripheral blood mononuclear cells (PBMCs) play important roles in the pathogenesis of IgA nephropathy (IgAN). Our study aimed to provide a deep understanding of IgAN and focused on the dysregulation of hsa‐miR‐590‐3p and its target gene HMGB2 in PBMCs. Three gene expression profile datasets (GSE14795, GSE73953 and GSE25590) were downloaded from the GEO database. The DEGs (differentially expressed genes)‐miRNA network that was associated with IgAN was constructed by Cytoscape, and HMGB2 and hsa‐miR‐590‐3p were selected for further exploration. The dual‐luciferase reporter system was utilized to verify their interaction. Then, the expression levels of HMGB2 and hsa‐miR‐590‐3p in PBMCs were detected by qPCR in another cohort, and the correlation of their expression levels with the clinical pathological manifestations and serum Gd‐IgA1(galactose‐deficient IgA1) levels was also investigated. HMGB2 was identified as the target gene of hsa‐miR‐590‐3p. Furtherly, the elderly patients had higher HMGB2 expression levels than the expression levels of the younger patients. As the serum creatinine, serum BUN levels increased, the expression of HMGB2 decreased; Besides, the HMGB2 expression was positively correlated with serum complement 3(C3) levels, and it also had a negative correlation with the diastolic blood pressure, but not reach statistical significance. What is more, both hsa‐miR‐590‐3p and HMGB2 expression had a slight correlation tendency with serum Gd‐IgA1 levels in the whole population. In conclusion, HMGB2, the target gene of hsa‐miR‐590‐3p, was identified to correlate with the severity of IgAN, and this provides more clues for the pathogenesis of IgAN.

and genetic effects contribute to it. Recently, the multihit pathogenesis model of IgAN has been widely accepted, 3 which indicates that circulating galactose-deficient IgA1 is the cause. The contribution of galactose-deficient IgA1 to IgAN pathogenesis has been validated in many studies. [4][5][6] Moreover, the presence of galactose-deficient IgA1(Gd-IgA1) in the glomerular deposits of patients with IgAN has been proven by immunohistochemical staining using the galactose-deficient IgA1-specific monoclonal antibody KM55, 7,8 which reinforces the important role that galactose-deficient IgA1 plays in IgAN.
Previous studies have shown that the aberrant deposition of glycosylated IgA1 in the renal mesangial area was from circulation. 9,10 The production of IgA1, including the O-glycosylation status of IgA1, was regulated not only by B cells but also by some cytokines secreted by T cells, dendritic cells, and monocytes, 11,12 all of which make up the majority of peripheral blood mononuclear cells (PBMCs).
Moreover, previous studies have identified many key proteins that are involved in the O-glycosylation of IgA1, including APRIL and BAFF, and all of these proteins were expressed in PBMCs, [13][14][15] indicating that PBMCs are a whole cell population and that further investigation is needed.
MicroRNAs (miRNAs) are a class of single-stranded, short RNA molecules that down-regulate gene expression by binding to specific sites within the 3′ untranslated regions (UTRs) of mRNAs to promote mRNA degradation or to interrupt translation processes. 16 MiRNAs can exist in the cell or can be secreted selectively out of the cell, with the regulatory function of gene expression and cell-to-cell communication. In recent years, great progress regarding miRNAs has been achieved in the field of nephrology. Many differentially expressed miRNAs in several kinds of human samples were identified as biomarkers or participants in IgAN pathogenesis, [17][18][19][20][21][22] indicating the important pathophysiological role of miRNAs in IgAN.
In the present study, we summarized and reanalysed the previously reported microarray data of PBMCs in IgAN to explore the miRNAs and target genes that are associated with IgAN.

| Microarray data preprocessing
All microarray data derived from PBMCs in IgAN were searched in the Gene Expression Omnibus database (http://www.ncbi.nlm.nih. gov/geo/). We obtained three datasets, including two mRNA arrays, GSE14795 23 and GSE73953, 24 and one miRNA array, GSE25590. 20 The data of patients with IgAN and controls were extracted in GSE14795 and GSE73953 for subsequent analysis (the data analysis pipeline is shown in Figure 1). The samples included in each dataset and the corresponding annotations for the array platform are listed in Table 1. All data were Log 2 transformed to achieve normality. In addition, data normalization was performed with the linear models for the microarray data (limma, http://www.R-proje ct.org) package in R. Principal component analysis (PCA) and clustering were also performed for the data quality control. The samples that did not reach the quality control standards were excluded, as shown in Table 1.

| DEG analysis and functional enrichment analysis
First, the samples in GSE14795 were used as the discovery cohort, and GSE73953 was used as the validation cohort to obtain the shared differentially expressed genes (DEGs). Second, the target genes of the differentially expressed miRNAs of GSE25590 were predicted using 4 miRNA databases:MiRDB , 25 Tarbase, 26 miRTarBase, 27 and TargetScan. 28 Finally, we merged the above results together and obtained the gene-miRNA network. QuickGO, 29 a web-based tool for gene ontology searching, was utilized for identifying the enriched functions in 'shared DEGs', and the Kyoto Encyclopedia of Genes and Genomes (KEGG), 30 a reference resource for gene and protein annotation, was used to assign 'shared DEGs' to specific pathways. P < .05 was used as the threshold value.

| IgAN-associated gene-miRNA network analysis
The Search Tool for the Retrieval of Interacting Genes (STRING) database 31 (http://string-db.org) was used to provide information on the different proteins, including the predicted and experimental interactions and the direct (physical) and indirect (functional) interactions of the proteins. The 'IgAN-associated genes and relative miRNAs' F I G U R E 1 Flowchart of the analysis process were mapped into the network. Cytoscape software, 32 a software for the integrated models of biomolecular interaction networks, was used to construct the IgAN-associated gene-miRNA network.

| Luciferase reporter assays
HMGB2 and hsa-miR-590-3p were chosen for further validation of the IgAN-associated gene-miRNA network. The dual-luciferase reporter system was applied to verify the interaction between hsa-miR-590-3p and its target gene HMGB2. Briefly, the 3′UTR sequences of the HMGB2 gene were amplified from genomic DNA and were subcloned directly downstream of the Renilla luciferase gene in the pmiR-GLO vector. A mutant version of the 3′UTR sequences of HMGB2 in the 'seed region' was also generated as the mutant control. Both constructs were verified by DNA sequencing.

| Detection of the expression of HMGB2 and hsa-miR-590-3p
Peripheral blood mononuclear cells (PBMCs) were prepared by density gradient centrifugation performed with Ficoll-Paque Plus (GE). After cell isolation, the total RNA was extracted using the commercial TRIzol Reagent (Invitrogen). The cDNA for miRNA and mRNA detection was synthesized from 1 μg total RNA using the Reverse Transcription System (for miRNA: TIANGEN, Beijing, China; for mRNA: Promega, Wisconsin, USA) and was stored at −20°C for the following amplification. The expression levels of HMGB2 and hsa-miR-590-3p were measured by semiquantitative reverse transcription-PCR performed with AceQ ® qPCR SYBR ® Green Master Mix (Takara), and this experiment was performed on an Applied Biosystem 7500 Real-Time PCR System (the primers are shown in Table 2). U6 snRNA and GAPDH were used for normalization. The expression fold change between the patients and the controls was expressed by the 2 −ΔΔCT method.

| Detection of plasma Gd-IgA1 levels
Plasma Gd-IgA1 levels were detected performed with a commercial enzyme-linked immunosorbent assay (ELISA) kit according to the manufacturer's specifications (IBL).  were obtained from GSE14795 and GSE73953, respectively. Then, 129 'shared DEGs' were obtained; finally, 38 differentially expressed miRNAs in the GSE25590 dataset were merged with 129 'shared DEGs' from GSE14795 and GSE73953, and 7 miRNAs and 19 target genes were identified (as shown in Table 3). Then, the genes were constructed with functional and pathway enrichment analysis by QuickGO and KEGG (shown in Table 4). We set the gene number above 6 for each GO term and a p-value less than 0.01 as the cutoff values, and we obtained 25 GO terms. Regarding the KEGG pathway, 2 pathways were identified with P < .05, but each included only 1 gene.

| Screened miRNAs and target genes associated with IgAN
IgAN is a multifactorial disease with unclear pathogenesis.
Previous studies have demonstrated that immune system disorders, susceptible genes and inflammation are the contributing factors.
The IgAN-associated gene-miRNA network was constructed using Cytoscape (shown in Figure 2). In total, five enriched pathways

| HMGB2 is a direct target of hsa-miR-590-3p
The 3'UTR of HMGB2 was predicted to have a conserved binding site for hsa-miR-590-3p (363th-370th bp), as shown in Figure 3A. To further confirm HMGB2 as the putative target of hsa-miR-590-3p, an hsa-miR-590-3p mimic or a negative control (NC) sequence was cotransfected with constructs containing the wild-type or mutant HMGB2 3'UTR into HEK293T cells. As shown in Figure 3B, the HEK293T cells that were cotransfected with hsa-miR-590-3p mimics

| Increased hsa-miR-590-3p expression and decreased HMGB2 expression in IgAN
To validate the contribution of HMGB2 and has-miR-590-3p in IgAN, 37 patients with IgAN and 9 healthy controls were TA B L E 2 The primer pairs for HMGB2, GAPDH and has-mir-590-3p   Figure 4B). Moreover, we also found a significant negative correlation between the expression of HMGB2 and that of hsa-miR-590-3p in the whole population of IgAN patients and healthy controls (r = −0.386, P = .008, shown in Figure 4C).  Figure 4D).

| HMGB2 was correlated with the severity of IgAN
After the validation of the decreased expression of HMGB2 in IgAN patients, which was down-regulated by the increased has-miR-590-3p levels, we further explored their association with clinical findings and the pathological lesions in patients with IgAN. In the patients with IgAN, the expression of HMGB2 showed a significantly positive correlation with age and serum C3 levels (age: r = 0.336, P = .042; serum C3: r = 0.416, P = .020; Figure 5A

| D ISCUSS I ON
IgAN is a complex multifactorial disease with an unclear pathogenic mechanism. For the first time, we integrated the published F I G U R E 2 IgAN-associated gene-miRNA network microarray data from PBMCs to find more clues to uncover the pathogenesis of IgAN.
In the present study, a total of 19 DEGs (all down-regulated) were identified through the comparison between the IgAN patients and the healthy controls in GEO datasets. The DEGs and differentially expressed miRNAs were mainly mapped to 5 IgANassociated terms (Figure 2). Among them, 'defense to response to bacteria' and 'cell surface receptor signaling pathway' were downregulated; however, 'inflammatory response to antigenic stimulus', 'protein N-linked glycosylation' and 'somatic diversification of immune receptors' were up-regulated. These results suggest that the immune system, inflammatory response and N-glycosylation modifications were associated with IgAN. As seen in patients with IgA nephropathy, the external defense system to bacteria was weakened; on the other hand, once stimulated by an antigen, the responses of the immune system and the inflammatory system were enhanced, which was consistent with the previously reported results of the genome-wide association studies (GWAS) of IgAN. 35,36 Moreover, although O-linked glycosylation has been widely accepted to be a key step in the initiation of IgAN, a few studies have indicated that N-linked glycosylation is also an important factor in the biological properties of IgA1. 37,38 There were a total of 6 genes and 2 miRNAs in the 5 terms. Both GCC2-and RAB1A-encoded proteins are required for transport from endosomes to the Golgi. 39 Moreover, RAB1A was reported to act as an oncogene to regulate cellular proliferation, growth, invasion and metastasis via the activation of the mTORC1 pathway in triplenegative breast cancer. 40 MAN1A1 encodes a class I mammalian Golgi 1,2-mannosidase to catalyze the hydrolysis of three-terminal mannose residues from peptide-bound Man(9)-GlcNAc (2) oligosaccharides. 41 The NAMPT-encoded protein is thought to be involved in many important biological processes, including metabolism, stress response and aging. 42 The protein encoded by RASA1 is associated with cellular proliferation and differentiation. 43 HMGB2, which has a high degree of similarity to HMGB1, was reported to be associated with cell viability, invasion and the chemotherapy resistance of glioblastoma and was reported to have antimicrobial activity. 44,45 Little is known about hsa-miR-648, and hsa-miR-590-3p was reported to be related to Lynch syndrome in a published research paper. 46 From the IgAN-associated DEG-miRNA network, it is clear that HMGB2 was in the core of the regulation network; the interaction between hsa-590-3p and HMGB2 was verified by the luciferase reporter assay and was validated in our IgAN cohort, in which miRNAhsa-590-3p and HMGB2 showed a significant negative correlation.
In fact, we also explored the HMGB2 protein levels in 5 patients with IgAN and 7 healthy controls, and found the HMGB2 expression levels in PBMCs were significantly lower in patients than healthy  controls, which was consistent with our previous results ( Figure   S1A, 1B). These findings suggest that HMGB2, which is targeted by hsa-miR-590-3p, may be associated with the pathogenesis of IgA which provides more clues about the pathogenesis of IgAN.

ACK N OWLED G EM ENTS
The work was supported by the Natural Science Foundation for

CO N FLI C T O F I NTE R E S T
The authors declared that there is no conflict of interest regarding the publication of this paper.

DATA AVA I L A B I L I T Y S TAT E M E N T
Raw data used during the current study are available from the corresponding author on reasonable request for non-commercial use.