The landscape of immune cell infiltration in the glomerulus of diabetic nephropathy: evidence based on bioinformatics

Diabetic nephropathy (DN) is an extremely common complication of diabetes mellitus, affecting up to 30% of individuals with type 1 diabetes mellitus and 40% of individuals with type 2 diabetes mellitus [1]. Over the last two decades, no new drugs have been approved to specifically prevent DN or to improve kidney function [2]. Current strategies to treat this disease mainly focus on intensification control of glycemic, blood pressure and other risk factors [3]. But a significant proportion of patients still progress to end stage renal failure overtime with a well control of risk factors, which highlight the urgent need to identify effective therapies.

Recently, DN is increasingly considered as an inflammatory disease, with immune modulation being involved in every stage. Immune cells infiltrated in kidney tissues, such as macrophages, T cells, and mast cells produce various inflammatory cytokines, metalloproteinases, and growth factors, which modulate the local response and increase inflammation [4‐8].

Although there is substantial evidence for the vital roles of macrophages, T cells, B cells, and mast cells, the effect of other immune cells including neutrophils, NK cells, and dendritic cells are less clear. Furthermore, the immune system contains various types of immune cell, which constitute a complex regulatory network, potential intercorrelations of them in the pathological process of DN are less clear. Therefore, an integrated analysis of immune infiltration, differentially expressed genes, and aberrant pathways activation in kidney tissues is essential.

Here, we used the CIBERSORT algorithm to analyze the abundance of 22 kinds of immune cells in the glomerulus samples of DN based on three microarray datasets mined from the Gene Expression Omnibus (GEO) [9]. Then an integrated bioinformatics analysis was conducted to screen immune-related differentially expressed genes(IR-DEGs)and explore aberrant pathways activation. Besides, we evaluated the correlation between IR-DEGs with immune infiltration cells and clinical indicator. Our research provides a potential value for immunotherapy of DN.

With the advancement of bioinformatics, efficient algorithms have accelerated the transition of various omics big data to new therapeutic targets. As one of the most robust algorithms for analyzing immune cell infiltration—CIBERSORT—has been used in multiple tissues [13‐16]. In this study, we used it to explore the landscape of immune infiltration in the glomerulus of DN.

In summary, we found that a variety types of infiltrated immune cell have altered during the pathogenesis of DN. Correlation analysis shows that the immune response may act as a complicated and tightly regulated network. In addition, we also screened out 6 immune-related hub genes which not only have correlations with differently infiltrated immune cells but also have a medium to strong correlation with eGFR (the most important renal function indicators).

Serpin Family A Member 3 (SERPINA3) was revealed to associated with acute immune responses in human renal allograft, and could be a novel biomarker for timely diagnosis [17, 18]. And Andrea Snchez-Navarro found urinary SerpinA3 could detect renal fibrosis and inflammation, with a particular potential for the early detection of AKI to CKD transition [19]. Lactotransferrin (LTF) was reported could be delivered to the Intracerebral hemorrhage affected brain by infiltrating polymorphonuclear neutrophils, and may assist in hematoma detoxification [20]. Qi Zhao found LTF could regulate the immune microenvironment of prostate cancer through JAK/STAT3 Pathway [21]. Nick Wlazlo revealed that Complement C3 (C3) were longitudinally associated with insulin resistance in type 2 diabetes over a 7-year follow-up period, and may reflect progression of metabolic dysregulation [22]. A diabetes control and complications Trial found PTGDS (prostaglandin-H2 D-isomerase), strongly and positively associated with the albumin excretion rate in Type 1 Diabetes Mellitus [23]. Boris B. Betz performed urinary peptidomics analysis in a rodent DN model (Cyp1a1mRen2, renin-dependent hypertension and diabetes mellitus synergized to produce massive albuminuria), and found that urinary epidermal growth factor (EGF) was more than twofold reduced in rats with DN in comparison with controls [24]. Hypochlorite modified albumin (ALB) is an active compound that is formed during the reaction between proteins, and was found to be much higher in concentration and to act as a mediator of oxidative stress and inflammation in patients with uremia or diabetes mellitus [25]. These results show that the 6 hub genes might act as vital roles in the pathological of DN.

It is of interest that neutrophils decreased drastically in DN patients, which is in contrast to the study of an STZ (streptozotocin) induced diabetic mouse model [26]. There are several possible reasons for this opposite conclusion: (1) the impact of intervention drugs may impair neutrophils activity. SCIENCE TRANSLATIONAL MEDICINE reported that angiotensin-converting enzyme inhibitors (ACEI) intervention significantly reduced neutrophils activity compared to treated with an (angiotensin receptor blocker) ARB or with no drug in vitro and in vivo. What's more, ACE knockout neutrophils showed decreased survival signaling and increased apoptosis [27]. (2) the negative feedback regulation of the immune microenvironment; (3) the sample size of this study is still not enough, resulting in abnormal results.

There are some related bioinformatics researches based on DN transcriptome data, but these literatures mainly focus on screening DEGs and exploring related pathways [28‐31]. To the best of our knowledge, there is no research report about using transcriptome data to reveal the landscape of immune cell infiltration in DN renal tissue. In addition, the hub genes we reported here were not only highly correlated with renal function, but more importantly, were highly correlated with differentially infiltrated immune cells. In short, we focused on immune infiltration and screened out 6 hub genes may associate with DN immune infiltration.

Nevertheless, some limitations still existed in our study: first, the infiltration of immune cells was estimated by algorithm, rather than based on more direct evidence, such as cell markers. Second, although batch effect removal and normalization were done, the homogeneity of samples between different datasets is still not fully guaranteed. For example, the DN samples may come from patients with very different pathological stages, or with other co-morbidities. And some NC samples come from tumor nephrectomies, which could have changes related to inflammation or alterations in vasculature. So, the results of immune cells differential infiltration stated here is far from definitive. Third, the correlation analysis of hub genes with eGFR is done based on a relatively small sample data (only 22 samples), so this may lead to over- or underestimation of its importance. And the exact mechanism of hub genes involved in immune infiltration had not been investigated through biological experiment in the present study. So, there is an urgent need to conduct in vivo experiments to confirm the results with higher-level evidence.

In conclusion, our work provides a global picture of the immune environment in the glomerulus of DN. The potential therapeutic targets identified in the present study may provide new insightful for clinical treatment.