A clinical and in-silico analysis of hsa-miR-21 and Growth Differentiation factor-15 expression in Diabetic Nephropathy

Riddhi Girdhar Agarwal All India Institute of Medical Sciences Jodhpur Purvi Purohit (  dr.purvipurohit@gmail.com ) All India Institute of Medical Sciences Jodphur https://orcid.org/0000-0001-8559-2911 Manoj Khokhar All India Institute of Medical Sciences jodhpur Anupama Modi All India Institute of Medical Sciences Jodhpur Nitin Kumar Bajpai All India Institute of Medical Sciences Jodhpur Gopal Krishna Bohra All India Institute of Medical Sciences Jodhpur Mahendra Kumar Garg All India Institute of Medical Sciences Jodhpur Praveen Sharma All India Institute of Medical Sciences Jodhpur


Introduction
Diabetes is a chronic metabolic disease characterized by hyperglycemia caused due to insulin insu ciency in the body. The prevalence of diabetes is highest in China and India, together accounting for nearly 180 million diabetic adults. The data is so alarming that the global prevalence of diabetes is anticipated to exceed 700 million by 2025 [1]. Type 2 Diabetes mellitus (T2DM) is a complex, multifactorial, and mostly preventable disease [2]. It can progress into micro and macro-vascular complications. One such complication isdiabetic nephropathy (DN), mainly characterized by albuminuria (≥300mg/day) and reduced glomerular ltration rate (GFR) [3].
At the molecular level, DN develops as a result of metabolic interactions, which activate intracellular signalling pathways, leading to extracellular matrix (ECM) protein accumulation, vessel permeability alteration, and proteinuria [4]. In DN, the hyper-glycaemic condition is associated with the upregulation of glucose transport-1 that also causes overexpression of transforming growth factor-β (TGF-β) in mesangial tubular cells or in ltrating renal cells [5]. This over-expressed TGF-β has been recognised as a key determinant of progressive renal function loss [6]. Furthermore, along with inducing the expression of TGF-β in mesangial cells, these factors also accelerate mesangial cell hypertrophy and ECM accumulation; as a consequence, GFR decreases and engenders chronic renal failure [6]. GDF-15, a member of the TGF-β superfamily, has been found to be associated with T2DM. Further, an elevated serum level of GDF-15 has been reported as a clinically valuable marker for predicting progression of diabetic kidney disease (DKD) and for individual risk stratification in DN patients, with normoalbuminuria and microalbuminuria [7]. Moreover, higher levels of GDF-15 corresponds to faster deterioration of kidney function and morbidity in DN patients [8]. Hence, it is important to validate GDF-15 as an individual biomarker for the progression of DN in humans. Several ndings have also highlighted that plasma GDF-15 level increases with the Mogensen stage, and is signi cantly correlated with microalbumin (mAlb) and eGFR, suggesting its value in early diagnosis, evaluation and prediction of the outcomes of DN [9]. A series of miRNAs, a class of short non-coding RNAs that regulate gene ex pression at the post-transcriptional level by binding to the 3'-untranslated regions of their target mRNAs [10], play a critical role in the regulation of many cellular and physiological activities [11]. These miRNAs either get up or downregulated in different kidney diseases [12], and can therefore serve as potential biomarkers in targeted therapy protocols [13].
In a study done on β cells, miR-21 inhibition was found to maintain the sugar level in patients with diabetic complications [5]. The expression of miR-21 is also known to be regulated by TGF-β1 and has been broadly studied in cancer biology because of its antiapoptotic effects and critical role in tumorigenesis [14]. Further, in human kidney biopsies, miR-21 was reported to increase phosphatase and tensin homolog (PTEN), and miR-21 antagonism resulted in decreased interstitial brosis, podocyte loss, in ammatory gene expression, albuminuria, macrophage in ltration and mesangial cell expansion [15]. In our previous review article, we have also highlighted the role of PTEN in the regulation of renal tubulointerstitial brosis and epithelial-mesenchymal transition (EMT) [16]. miR21 levels have also been reported to elevate with increasing disease pathology in human renal biopsies and in experimental mouse models of early/late DN [17]. However, miR-21 is regulating which member of TGF-β family, has not been investigated earlier in diabetes. Therefore, based upon previous literature we anticipated a common link between miR-21 and TGF--β family member GDF-15, therefore we investigated their interaction using bioinformatics and computational analysis, that has emerged as an e cient tool to study pathways involved in differentially expressed genes, protein-protein interaction (PPI) networks and to predict miRNA-mRNA targets [18]. Further, the expression level of serum GDF-15 was investigated in T2DM DN patients and healthy controls and was correlated with various biochemical and anthropometric parameters of the study population. Additionally, we also analysed the levels of circulating hsa-miR-21 and its correlation with serum GDF-15. Our results support the potential of a raised serum GDF-15 and circulating miR-21 to serve as therapeutic targets of progressive renal deterioration. The interconnecting networks between GDF-15 and its nearest neighbours were investigated in diabetes using Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) [19]. The protein IDs of insulin and GDF-15 were submitted to STRING to obtain functional interpretations of their targets in Homo sapiens ( Figure 1). Kyoto encyclopedia of genes and genomes (KEGG) components were analysed to obtain pathway enrichment by STRING statistics taking con dence of 0.4, and the signi cantly enriched pathways were identi ed based on a value of P < 0.05. Plug-in of Cytoscape tool was employed [20] for cluster formation and molecular complex detection (MCODE).

Functional annotation and enrichment analyses
The functional features of all the genes related to T2DM involved in PPI networks were analysed by two different web servers DAVID (Database for Annotation, Visualization and Integrated Discovery) [24] and STRING Version 11.0 [19]. DAVID was used for functional enrichment such as KEGG pathways and signi cance of gene ontology (GO) to assess the biological process, cellular components and molecular functions. Signi cantly enriched GO terms and KEGG pathways were selected based on p<0.05 and further, a false discovery rate of <0.05 to avoid false positives.

Construction of MicroRNA-mRNA target regulatory network
Common transcription factors were utilized to construct PPI networks, and miRNet Version: 2, a wellestablished miRNA target prediction database was used to predict the target miRNA of genes involved in the PPI networks [22,25].
Considering the literature search and predicted miRs against targeted transcription factors, miR-21 was scrutinized. Further, Targetscan (http: www.targetscan.org) was used to predict the target mRNA according to miR-21 seed region. A heatmap of genes and proteins involved in PPI Networks, representing their expression in various tissues, was obtained using FunRich (Functional Enrichment analysis tool) [26].

Study population
Thirty DN patients visiting the outpatient department (OPD) of the Department of Nephrology, All India Institute of Medical Sciences (AIIMS) Jodhpur, and thirty healthy volunteers working as healthcare professionals were recruited in the study. The inclusion criteria for patients was HbA1c > 6.5% and urinary mAlb >300 mg/dl. Inclusion criteria for controls were non-diabetic subjects without any in ammatory disease or history of renal disorder.
Since GDF-15 gets upregulated due to damaged vascular endothelium, patients diagnosed with type 1 diabetes mellitus (T1DM), hypothyroidism, coronary artery disease, chronic obstructive pulmonary disease, malignancies, hepatic disorders, rheumatoid arthritis and thalassemia were excluded. In hemochromatosis, GDF-15 is secreted by erythroblasts [27], and in pregnancy, GDF-15 levels are raised by the placenta, therefore these patients were also excluded.
Venous blood samples of the study population were collected under the ethical approval of the Institutional ethics committee (IEC) of AIIMS, Jodhpur to obtain whole blood, serum and plasma. Informed consent was taken from all the participants. Urine samples were collected from the study subjects for the estimation of urinary mAlb.

Study protocol
On the same day of blood sampling and handing out of 24hr urine samples, anthropometric measurements of all the participants were recorded. Height (by a stadiometer), weight (by a weighing machine), and waist-and hip circumference (by non-elastic measuring tape) were measured. These measurements were then used to calculate body mass index (BMI) and waist-hip ratio (WHR). Systolic blood pressure (SBP) and diastolic blood pressure (DBP) was recorded using a sphygmomanometer.

Biochemical analysis
Enzymatic colour test of human serum was performed on Beckman Coulter AU analyser using End-point method, for the quantitative determination of Fasting Blood Sugar (FBS), Low-Density Lipoprotein (LDL), High-Density Lipoprotein (HDL), cholesterol, urea and creatinine.. The Serum electrolytes sodium, potassium and chloride were analysed on the Beckman Coulter AU analyser using the Ion-Selective Electrode (ISE) principle. HbA1c was measured by latex agglutination inhibition assay. Serum insulin was analysed using the fully automated chemiluminescence analyser Diasorin Liasion. Urinary mAlb was estimated by immunoprecipitation method. The GFR of the study population was calculated using the MDRD formula GFR (mL/min/1.73 m²) = 175 × (serum creatinine in mg/dl) -1.154 × (Age) -0.203 × (0.742 in females).

Analysis of serum GDF-15
Thermo Scienti c Pierce Human GDF-15 ELISA kit was used to quantify human GDF-15 in serum, according to manufacturer's instructions. The concentration of the analyte (GDF-15) was used to obtain a standard curve. The concentration of serum GDF-15 in the samples was analysed using the standard curve by BioTek Gen5 Data analysis software.

Analysis of circulating hsa-miR-21
Whole blood samples of the study population were used to isolate total RNA using RBC lysis buffer in a ratio of 1:3 for Blood: RBC lysis buffer. The sample was centrifuged at 1300 rpm at 4˚C for 15 minutes and 1 ml PBS was added to dissolve the pellet, followed by centrifugation at 3000 rpm for 2 minutes. Trizol was added to the pellet (in a ratio of 10:1 for blood: trizol) to obtain a homogenous solution. Isopropanol was used to remove the aqueous layer (in a ratio of 1:1 for aqueous layer: isopropanol) and the contents were brie y vortexed. The pellet was then dissolved in 1 mL of 1% ethanol and centrifuged at 12000 x g for 5 mins. 20 µL of DEPC treated water was added after air-drying the pellet followed by incubation at 65˚C for 3-4 minutes. The sample was stored at -80˚C till further processing. Thermoscienti c NanoDrop One c was used to quantify the extracted total human RNA. The absorption maxima of RNA lies at 260 nm, and DEPC treated water was used as blank. An absorbance of ~2.0 at a ratio of 260 nm and 280 nm was accepted as "pure" for RNA.

Statistical Analysis
IBM SPSS Statistics 21.0 for Windows (SPSS Inc, Chicago, IL) was used to analyse the skewness and kurtosis of various study parameters. Mann-Whitney U test was used to test the research hypothesis and Spearman correlation and logistic regression analysis were used to evaluate the relation between variables. Microsoft Excel software was used for graphical representation of the results.

GO enrichment analysis
Classi cation of genes, involved in PPI network, into the following three classes, was used for GO analysis ( Figure 2): Cellular component class: the genes were enriched for the cellular components like INSR complex, protein-containing complex, nuclear transcription factor complex, receptor complex and plasma membrane protein complex.
Biological Process (BP) Class: The proteins were enriched in the cellular response to organonitrogen, peptide, response to insulin, regulation of transferase, kinase, phosphorylase activity of proteins, signal transduction, cellular response to growth factor, and signalling pathways of insulin receptor, phosphatidylinositol 3-kinase (PI3K), IGFR and mitogen-activated protein kinase (MAPK).
Molecular Function (MF) class: The proteins were enriched in the activation of INSR binding, PI3K binding, IGFR binding, RNA polymerase II regulatory region, DNA-binding transcription activator activity, RNA polymerase II-speci c insulin binding, IRS binding, PI3K activity, transmembrane receptor protein tyrosine kinase activity, phosphatidylinositol-4,5-bisphosphate 3-kinase activity.

Prediction of transcription factor and construction of miRNA regulatory networks
The target miRNAs of genes of interests were predicted by miRNet Version: 2 (an integrated platform linking miRNAs, targets & functions) and miRDB. The regulatory relationship of identi ed miR-21 with GDF-15 and its transcription factor was investigated by miRNet and miRDB. Cytoscape was used to construct networks based on the correlation between the target genes and their regulating transcriptome. The identi ed genes were ATF3, CEBPB, EGR1, KLF4, PPARG, TP53 and TWIST2, which can regulate the levels of GDF-15. In the metabolic regulatory process of T2DM and insulin resistance, IRS1, IRS2, INSR, IGF1R, INS, AKT1, PPARG, CEBPB, EGR1, TP53, KLF4, ATF3, GDF15, TWIST2 served as important nodes in the PPI networks. MicroRNA-21 was observed as a multi-targeting miRNA as it interacted with KLF4, TP-53, and CEBPB transcription factors, which were also linked with GDF-15. Additionally, hsa-miR-21 directly targeted IGF1R. Further, the interactions also suggested a correlation between GDF-15 and miR-21, supposedly through the common transcription factors.

Demographic characteristics of the study population
In the present study, 30 DN patients and 30 healthy controls were included. Among the patients, 70% were males and 30% were females (Table 1), which re ects a higher prevalence of DN in males in the study population. This higher frequency of DN in males is predominantly due to the cross-sectional nature of the study design and does not indicate a gender-based risk of males towards DN. The age distribution of DN ( Figure 2) revealed a near gaussian distribution of DN in the cases (40-80 years). The prevalence of DN in central age group (50-70 years) can be due to the fact that DN is a late onset disease, and usually advances with long standing diabetes, therefore a smaller number of patients belonged to the other age groups.

Anthropometric characteristics of the study population
A highly signi cant difference (p = 0.002) in the WHR was found between healthy controls and DN patients, wherein DN patients were found to be overweight (Table 1). DN patients also had a signi cantly high systolic and diastolic BP as compared to controls (Table 1). High BP of DN patients can be reasoned for hypertension (a common complication of diabetes) and a progressive cardiac myopathy condition, which is another micro-vascular complication associated with T2DM.

Clinical characteristics of the study population
The levels of FBS (184.49 ± 49.05 mg/dL), insulin (27.02 ± 28.49 µIU/mL), HbA1c levels (8.25 ± 1.70%) and HOMA-IR (12.19 ± 2.81) were found to be signi cantly higher in DN patients than healthy controls (  An extremely signi cant difference was found in the values of FBS, insulin, HOMA-IR, urea, creatinine, Na, Cl, TGL, urinary microalbumin and GDF-15, with p ≤ 0.001 between DN patients and healthy controls. A highly signi cant difference in the value of HbA1c (p ≤ 0.05) was also found between the groups in the study population.

Analysis of miR-21 after quanti cation by RT-PCR
The FCE of circulating hsa-miR-21, assessed using RT-PCR in the study population, was found to be 9.18 folds higher in DN patients as compared to healthy controls (Table 3). Additionally, serum GDF-15 had a positive correlation withcirculating hsa-miR-21.

Discussion
At the molecular level, the factors that contribute to DN such as ECM protein accumulation, vessel permeability alteration, and proteinuria, develop as a result of interactions between various metabolic factors, which activate intracellular signalling pathways responsible for triggering in ammatory mediators and release of growth factors [5] like TGF-β1. The levels of TGF-β1 have been reported to increase under diabetic conditions in renal cells, including mesangial cells , hence TGF-β has been recognised as a key determinant of progressive renal function loss [6]. GDF-15, a member of the TGF-β superfamily, has been recently reported as an early marker of IR and mitochondrial dysfunction. An elevated serum level of GDF-15 was found to be associated with T2DM [7]. Several ndings have also highlighted that plasma GDF-15 level increases with the Mogensen stage as an independent risk factor for increased microalbuminuria in DN patients. GDF-15 is also signi cantly correlated with mAlb and eGFR, suggesting its value in early diagnosis, evaluation and prediction of the outcomes of DN. Furthermore, according to our in-silico ndings, and previous literature, TP53 and its family members, p63 and p73 have been reported as transcriptional regulators of GDF15, as its promoter region contains two p53-type response elements, RE1 and RE2, wherein RE2 confers p53-speci c transactivation [28]. KLF4 acts as a transcriptional repressor of p53 [29]. In ovarian cancer cells, the expression of CEBPB was positively correlated with GDF15 expression and CEBPB was identi ed to bind with GDF15 gene promoter through luciferase reporter assay. Thus, CEBPB caused a positive regulation of GDF15 expression in cancer cells through epigenetic modi cation [30]. In another study, the CpG locus in MIR21 promoter was observed to be a conserved binding site of transcription factors CEBPB, MEIS3, and TEAD4, which were co-expressed with miR-21 in tumors [31]. Therefore, KLF4, CEBPB and miR-21 jointly augment EMT via the AKT/ERK1/2 pathway by upregulating the levels of GDF-15 [32]. EMT plays an important role in renal interstitial brosis (RIF) with DN [33].
Our in-silico ndings have also suggested the association of insulin with GDF-15 and IGF1R, which implies the role of increased GDF-15 in insulin resistance along with endothelial dysfunction, metabolic derangement and in ammation as proposed in vascular injuries and cardiovascular complications elsewhere [34]. In another study, p53 dependent expression of GDF-15 has been suggested as the link between obesity and insulin resistance, wherein activation of p53 in adipose tissue led to increased production of proin ammatory cytokines, subsequently leading to insulin resistance, and diabetes [34,35]. Therefore, the in-silico ndings of this study reiterate the connection of GDF-15 with insulin in diabetes, and along with the previous literature highlights the likelihood of GDF-15 mediated microvascular complications in DN. In a previous nding, diabetic dyslipidaemia has been reported to promote progression of DN, with few studies suggesting that TG-rich lipoprotein particles predominantly containing apolipoproteins (apos) E, C and B serve as major promoters of DN. Further, in DN patients, plasma TG levels were also suggested to be high, due to the reduced activity of hepatic lipase (HL) and lipoprotein lipase (LPL), which hydrolyze TG [35]. This is in accordance with our nding of elevated TGL levels in the patient population. The reduced serum Cl levels observed in the current study may be due to elevated serum ketoacids as observed by Yasuda K et al., wherein they have reported that in diabetics, hypochloremia can result due to elevated serum ketoacids and the ratio of Cl/ Na in DN patients was found to be signi cantly lower in their study as well [36]. In our molecular analysis, we found the serum levels of GDF-15 (Table 1) to be nearly ten (10) folds higher in DN patients (median 5507 pg/ml) as compared to healthy controls (567 pg/ml), (p<0.0001) ( Figure 4A). Further, GDF-15 levels were also observed to increase with stages of DN. The highest level corresponded to Stage V (ERSD), thus con rming increasing GDF-15 to be associated with advancing stages of DN.
The levels of miR-21 have also been reported to be higher in human renal biopsies and in experimental mouse models of early/late DN and non-DN, further, renal miR-21 knockdown showed downregulation of TGF-β1 signalling in a mouse model of T2DM [37]. The increased miR-21 expression has also been identi ed in the kidney biopsy samples of renal transplant patients with brotic kidney disease and in the urine of brotic patients with IgA nephropathy [38]. Since circulating miRNAs can serve as a non-invasive marker of any disease condition, Zununi and his colleagues assessed plasma samples for circulating miRNAs and observed a correlation between expression levels of miR-21 in plasma and serum creatinine (r =-0.432, P =0.03) in renal transplant patients [39]. In another study, circulating miR-21-5p has been reported to be closely associated with endothelial dysfunction and in ammation, which are two characteristic developmentsof T2DM progression. However, there are no reports on the FCE of circulating miR-21 in relation to DN. In the present study, we observed FCE of circulating miR-21 to increase up to 9.18 folds in DN patients in comparison to healthy controls ( Figure 4B). The levels of miR-21 were also found to be positively correlated with the levels of GDF-15, although the association was non-signi cant (which can be due to the small sample size). This positive correlation indicates the possibility of mechanistic interaction between miR-21 and GDF-15. However, GDF-15 upregulation can be reflected in any vascular endothelial damage, therefore, it can appear in diverse tissues damage responses apart from nephropathy, thus enhancing the levels of circulating GDF-15 [7]. Further, in a study done by Weronica et al., to investigate the role of GDF-15 in cardiovascular disease, GDF-15 and DNA methylation level of miR-21 promoter was found to be inversely correlated. They suggested that raised levels of GDF-15 possibly lead to demethylation of miR21 promoter, increasing its expression in cardio vascular disease [40]. However, this arena is still unexplored in DN, and a similar mechanism may be functional in DN.

Conclusion
This study, which is the rst to the best of our knowledge, has explored the role of circulating hsa-miR-21 in DN patients in relation to and

Declarations
Funding: This study was funded and supported by All India Institute of Medical Sciences, Jodhpur, India

Con ict of interests:
Authors have no con icts to report.
Data availability statement: The data will be provided when on a reasonable request.
Ethics approval: The study was approved by AIIMS Institutional Ethics committee prior to commencement.

Consent to participate
All participants were asked for an informed consent to prior to participation, with a free will to withdraw if they wished to. indicates a signi cant difference between the variables and p ≤ 0.005(**) indicates an extremely signi cant difference.  Figure 1 Flow Chart of the methodology.