Diabetic nephropathy (DN) is one of the common complications of diabetes all around the world and refers to the cases with biopsy proven kidney lesions.1 DN is known as the most common microvascular side effect of diabetes and approximately 50% of patients with DN ultimately need dialysis and kidney transplantation.2,3

DN is accompanied by some structural changes in the renal glomeruli which finally can lead to albuminuria.4,5 So far, various genetic and non-genetic factors were discovered to take part in the pathogenicity of DN.6 However, despite a huge number of investigations, the underlying molecular mechanisms of DN and the details of its pathogenicity are not yet fully understood. In addition, most current treatment options for DN are ineffective and include supportive and nonspecific treatments such as blood sugar control, blood pressure control, and dietary restrictions.7 Recently, thanks to the advent of high-throughput strategies, there has been a great opportunity for researchers to catch a big map of little changes in complicated diseases like DN. In this context, transcriptomic studies are good examples that enable researchers to assess a large number of expressional alterations in any disease state. Understanding the disease pathogenicity and discovery of potential therapeutic targets could be two big advantages of these omics-based investigations.8–10 However, transcriptomic studies are facing some well-known challenges. These studies have yielded a list of candidate genes, some of which could be false positives, then the achieved list might lack real-world genes of interest.11 The first consequence of this compromise is exploring unrelated gene ontologies or pathways, which finally give misleading information. To address this issue, one possible approach is to collect all the available and related data and perform a meta-analysis. Using this method could finally give more robust and consensus results.12 Accordingly, meta-analysis can greatly reduce the false positives caused by differences in sampling, homogeneity, and study design in transcriptomics datasets.13

The aim of the present study was to draw together all the publicly available DN-related glomerulus datasets and perform a meta-analysis in order to reach a consensus list of differentially expressed genes (DEGs) in this disease. Briefly, after selecting, quality checking, and normalizing, the related expression profiles were subjected to meta-analysis and the resulted list of candidate genes was considered for functional enrichment analyses. Besides, to identify the top molecules related to the pathogenicity of DN, a protein-protein interaction (PPI) network consisting of the candidate genes, as well as their related microRNAs (miRNAs) and transcription factors (TFs), was constructed and analyzed.

Despite extensive research dedicated to understanding the molecular mechanisms of DN, its pathogenicity is yet to be completely understood. This lack of knowledge might be due to the heterogeneous nature of the disease or the methodological insufficiencies of previous studies. Systems biology offers valuable approaches not only to decipher the underlying molecular mechanism of diseases but also to discover novel therapeutic targets for their possible control and prevention.

Meta-analysis of transcriptomics datasets is one of the efficient methods, which can provide more consensus results about heterogeneous diseases like DN. By integrating expression data acquired from independent studies, Meta-analysis can enhance the statistical power and robustness of a study. Consistency in the results of meta-analysis studies has made them suitable choices for predicting more reliable drug targets, and uncover more specific disease-related pathways.

So far, there have been several studies integrating DN transcriptomics datasets. However, to the best of our knowledge, there is no meta-analysis study conducted on human DN transcriptomics datasets. In one recent experiment, after an independent analysis of six DN-related datasets, regardless of the tissue type common DEGs were selected for the construction and analysis of a co-regulatory network.23 Integrating different sample types (blood and tissue biopsies) in this study might cause more heterogeneity and affect the overall result. In addition, the quality control and comparability of the analyzed datasets were not assessed in this study.

One of the forms of DKD is DN that is primarily initiated as a glomerular disease, therefore in order to understand the earliest transitions occurring, the tubular gene-expression profiles were excluded to create more homogenous data.24 Bearing in mind the differences in gene expression between tubular and glomerular tissues, to decrease potential heterogeneities, only microarray datasets containing glomerular expression profiles were selected.25

PCA is a valuable tool utilized to assess the quality of gene expression profiles. It is sensitive toward batch effects, poor-preprocessing analysis, and differences between conditions.26 We applied PCA to check the quality of each dataset and detect possible outliers. After normalization and batch effects removal steps, meta-DEGs were obtained by random effect size model.

Further on, we included other regulatory elements (miRNAs and TFs) and constructed an interactive regulatory network to emphasize their roles as upstream regulatory elements and insert more weight into the constructed network.27–29 TFs and miRNAs have long been recognized as two types of regulatory elements having a pivotal role in activating or suppressing gene expression. It has been shown that the interplay between these molecules regulates intra-module and inter-module pathways, and distinct niches in gene regulatory networks.30 The regulatory network was constructed by inserting these two groups as the main upstream regulators of gene expression. Several TF and miRNA molecules were recognized as the top upstream regulators based on their degree scores in the network (Table 4). SP1, STAT3, NFKB1, RELA, and E2F1 were the top 5 TFs with a potential role in the pathogenesis of DN. In case of miRNA molecules, hsa-mir-16-5p, hsa-mir-335-5p, hsa-mir-17-5p, hsa-mir-20a-5p, and hsa-mir-93-5p were the top miRNAs regulating meta-DEGs. Supplementary file 2 (S2) listed some investigations about the potential role of these regulatory hubs in kidney-related diseases.

Our enrichment analysis showed three of the top enriched pathways to be associated with the immune system (innate immune, adaptive immune, and immune system). Although metabolic and hemodynamic factors are traditionally thought of as the root of DN, different evidence have introduced immune mechanisms and inflammasomes as major pathogenic factors.31,32 Scores of the therapeutic approaches have been developed or repurposed to target the immune system molecules and inhibit the inflammation in DN disease.33 For instance, Baricitinib a selective inhibitor of JAK-1 and JAK-2, was shown to reduce albuminuria (up to 30%) and some inflammatory markers in DN patients.34 Back to the results of the present experiment, JUN, CDKN1B, VEGFA, PTEN, EGFR, MYC and TP53 were among the identified hub genes with regulatory roles in the immune pathway that could be considered for more investigations as therapeutic targets in DN.

The association of several of our top genes and DN have previously been studied, but as far as we know, there were no investigations on the possible linkage between other identified hubs and DN pathogenicity. For example, HMGA1 is known as a pro-oncogenic gene producing an architectural transcription factor that regulates genes in different cancer types.35 Its role in insulin resistance is under research, but HMGA1's role in kidney injury has yet to be studied.36CCND2, BTG2, E2F3, CDK2,SRF, and CCNB1 were other relatively novel hub genes that have not thoroughly been explored as potential effectors in DN. Therefore, detailed investigations are recommended.

In conclusion, the result of this meta-analysis experiment was a consensus list of differentially expressed genes in the pathogenesis of DN. Further on, we recognized the highly related regulatory elements including miRNAs and TFs, and eventually presented a more robust interactive network including all these elements. Altogether, the obtained list of molecules could be the focus of more investigations in order to understand the underlying aspects of the DN pathogenesis. Moreover, the recognized hub molecules either meta-DEGs or other regulatory elements including miRNAs and TFs in the DN-related network is valuable for further studies, especially in creating new treatments in the clinical context.

Amir Roointan and Fatemeh Hojjati participated in the design, screening, selection, meta-analysis, interpretation of data, and drafting of the manuscript. Alieh gholaminejad and Yasin Eshraghi were contributed to the meta-analysis, and interpretation of data. Yousof Gheisari contributed to the interpretation of the analyzed data and preparing the manuscript draft. All authors reviewed the manuscript.

Not applicable.

Not applicable.

The analyzed datasets by the present study are available in the GEO repository [https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE104948], [https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=gse1009], [https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=gse47183], [https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=gse96804], [https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=gse30528].

This work was supported by the Isfahan University of Medical Sciences [grant number 340027].

The authors declare that they have no competing interests.