Six children with nephrotic syndrome, 2 diagnosed as SRNS and 4 as SSNS were selected from an on-going study as per selection criteria (Supplementary Table 1). Two apparently healthy young adult volunteers without a recent febrile illness (before/within 2 weeks of obtaining blood sample), or hypertension or history of chronic or acute disease associated with inflammation were used as controls (Table 1). All the patients received 2-mg/kg/day prednisone treatment at initial episode of nephrotic syndrome, and were classified according to the standard guidelines [1]. Both the SRNS patients were tested and identified as mutation-free for the commonly mutated genes (NPHS2 and WT1). Among the 4 SSNS patients, 2 had infrequent relapses (infrequent relapse nephrotic syndrome, IFRNS) and 2 had frequent relapses (relapse nephrotic syndrome, FRNS) (Table 1). Both the IFRNS patients were not on any medication during sampling while both the FRNS patients had relapsed while on levamisole at the time of sampling. On follow-up, FRNS patients were treated with cyclophosphamide while the SRNS patients received tacrolimus and cyclophosphamide as steroid sparing alternatives (Table 1). Blood samples (5 mL) were collected from all 6 nephrotic patients at the time of relapse (urine albumin ≥3+). Blood from all subjects collected in ethylenediamine tetraacetic acid tubes, were spun at 4℃ @ 5,000 rpm for 15 minutes to separate plasma which was then stored at -80°C until further use. We also confirmed podocyte injury following exposure to plasma from children with SSNS and SRNS by assessing changes in podocyte morphology using immunofluorescence microscopy and scratch assay. The methods have been detailed in supplementary data.

Conditionally immortalized human podocyte cells (passage 22) obtained from Dr. Moin Saleem’s lab at University of Bristol, were used for these experiments. Cells were propagated at 33℃ until they were 70% confluent, and allowed to differentiate for 14 days at 37℃ in Rosewell Park Memorial Institute (RPMI) 1640 media with L-Glutamine and 25mM HEPES (N-2-hydroxyethylpiperazine-N-2-ethanesulfonic acid) supplemented with 10 % fetal bovine serum and 1% penicillin streptomycin with media change once in 2 days [10]. Differentiated cells were exposed to human plasma diluted to 5% with RPMI media (e.g., 1-mL culture medium contains 50 µL of human plasma + 950 µL of RPMI media) for 2 hours and total RNA was extracted using Trizol reagent (Invitrogen Life Technologies, Carlsbad, CA, USA) as per manufacturer’s instructions. RNA concentration was assessed using a NanoDrop 1000 spectrophotometer (ThermoFisher Scientific, Waltham, MA, USA) and integrity was analyzed using Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA) through RNA integrity number (RIN).

Microarray was performed using single color (CY3) Agilent Human Whole Genome 8x60k Gene Expression Array (AMADID: 39494) as per the manufacturer’s instructions. Gene expression data were extracted from images using Feature Extraction software v 10.7 of Agilent, pre-processed for background correction and data normalization between arrays using R software [11]. The normalized file was further filtered to remove pseudo genes and obtain average intensity values for duplicate probes. Genes expressed in the diseased samples (SSNS and SRNS) with standard deviation ≤0.1 compared to the healthy control were removed because of low variance to obtain a subset of data (16,355 genes) used for all further analysis. Unsupervised hierarchical cluster analysis was performed with ‘gplots package’ from R by using the Pearson similarity metric and Ward D2 as a clustering method [12].

Differential gene expression (DGE) analysis was performed using LIMMA package in R software in 2 ways – (1) diseased groups (SSNS or SRNS) versus healthy control and (2) between the diseased groups SSNS versus SRNS [13]. Genes with P<0.05 and -1≤ log FC ≥1 were filtered to identify significant differentially expressed genes (DEGs) with both the analyses (SSNS or SRNS vs. healthy control and SSNS vs. SRNS).

A Venn diagram was constructed with DEGs (SSNS or SRNS vs. healthy controls) which identified common and unique genes. These significant DEGs were annotated using PANTHER classification system (Protein Analysis THrough Evolutionary Relationships, http://pantherdb.org) [14]. To identify relevant pathways, web-based software WebGestalt (WEB based GEne Set AnaLysis Toolkit) and Enrichnet were used. Functional analysis of transcriptional networks and interactions between identified pathways were analyzed using Toppcluster and visualized using Cytoscape [15-18].

To identify intrinsic molecular subtypes and to visualize the gene expression pattern, a heatmap was constructed using corresponding expression values of DEGs (P<0.5) between SSNS and SRNS in R software. Further filtration criteria of (1< log FC >1) was applied to identify significant genes differentially regulated between SSNS and SRNS.

To validate the microarray results, significant DEGs representing common pathways and genes with prior reports as podocyte injury markers were selected as candidate genes. The primers were designed using Primer 3 plus (http://primer3plus.com/cgi-bin/dev/primer3plus.cgi) (Supplementary Table 2) [19]. Total RNA extracted from podocytes exposed to plasma of same subjects diluted to 5% using Qiagen RNeasy mini kit was converted to cDNA with 1 μg of RNA as input using the high capacity reverse transcriptase cDNA kit (ThermoFisher Scientific). Amplification was performed in duplicate using the Power SYBR Green PCR master mix (ThermoFisher Scientific) in StepOneplus real-time PCR machine (ThermoFisher Scientific). The data was normalized using GAPDH (glyceraldehyde-3-phosphate dehydrogenase) expression and analyzed according to the ∆∆Ct method [20].

All experiments were carried out according to protocols approved by Institutional Ethical Committee of St. John’s Medical College Hospital (approval number: 239/2015). Informed consent was obtained from all individual participants included in the study.

To determine whether nephrotic plasma (SSNS and SRNS) was able to trigger morphological changes in podocytes leading to cytoskeletal disruption and footprocess effacement, we performed morphology assay, immunofluorescence microscopy, and scratch assay. Podocytes treated with SSNS as well as SRNS plasma showed morphological changes viz. rounding up of cells due to retraction of foot processes as early as 6 hours (Supplementary Fig. 1). Immunoflurescence assay showed focal adhesion complexes stained with antivincullin antibody as green punctate marks. The cytoskeleton was visualized by phalloidin antibody as red parallel strands of actin fibers (Supplementary Fig. 2). In a healthy podcyte (untreated and incubated with healthy control), focal adhesion complexes and intact cytoskeleton were observed in all-time points (1, 3, 6, and 12 hours). However, Podocytes treated with SSNS and SRNS plasma showed loss of focal adhesion complexes and disorganization of cytoskeleton (loss of parallel arrangement of actin fibers) at 6 hours. Scratch assay also showed exposure of nephrotic plasma (SSNS and SRNS) to podocytes leads migration of podocytes (Supplementary Fig. 3). Thus, it is evident that nephrotic plasma consists of putative toxic factors capable of inducing podocyte injury.

Unsupervised hierarchical clustering of global gene expression revealed 2 major clusters with 6 subclusters (Fig. 1). Correlation with patient’s clinical data such as drug response and disease outcome revealed that major clusters correctly represented the SSNS and SRNS subgroups of NS while the sub clusters correlated with treatment response and clinical course. Within SSNS cluster, podocyte exposed to plasma of IFRNS patients (SSNS 2 and 3) formed a single subcluster and were separated from FRNS patients (SSNS 1 and 4). Among the 2 FRNS patients, SSNS 1 placed in outer branch had poor response to levamisole and oral cyclophosphamide but responded to tacrolimus, while the other child (SSNS 4) who had a good response to levamisole was placed in a separate branch. Within SRNS subcluster, a child who attained complete remission on tacrolimus was separated from the child with poor response to tacrolimus who progressed to end-stage renal disease. Although podocytes treated with controls grouped with those treated with the more aggressive phenotype, they formed a distinct subcluster.

As compared to podocytes treated with healthy plasma, a total of 105 (58 up-regulated and 47 down-regulated) and 139 genes (78 up-regulated and 61 down-regulated) were identified as significant in SSNS and SRNS treated podocytes respectively (Supplementary Table 3). Of these, 55 genes (40%) were common in both SSNS and SRNS (Fig. 2A, Supplementary Table 3). Among the 105 DEGs in SSNS, 50 (47.6%) were unique to SSNS, while out of 139 DEGs in SRNS group, 84 (60.4%) were unique to SRNS. Gene ontology (GO) analysis for biological processes of DEGs in SSNS as well as SRNS, revealed genes to be involved in cytokine and interferon-mediated pathways, in defense response to virus and transcription regulation pathways (Supplementary Table 4A, B). Molecular function analysis showed that DEGs were mostly involved in binding and regulation of transcription factors (Supplementary Table 5A, B). Pathway analysis revealed the DEGs to be mainly involved in mitogenactivated protein kinase (MAPK) signaling pathway (6 genes), transforming growth factor (TGF)-β signaling pathway (5 genes), and osteoclast differentiation (4 genes) (Table 2). EnrichNet identified TGF-β pathway (ID3, ID1, SMAD7, GDF6, TNF) and Janus kinase-signal transducer and activator of transcription (JAKSTAT) pathway (CSF3, CLCF1, IL11, SOCS3) as important pathways (with high kidney specificity score) modulated during podocyte injury (Table 2).

Network analysis of the enriched genes revealed tissue necrosis factor (TNF) as a significant gene involved in multiple pathways such as TGF-β signaling, cytokine and inflammatory, osteoclast differentiation, and MAPK signaling pathway. SOCS3 was another significant gene involved in osteoclast differentiation pathway and cytokine signaling (Fig. 2B). For validation of microarray data, significant DEGs representing multiple pathways (ID1, ID3, FOSB, SOCS3, SMAD7) were selected as candidates that are modulated during podocyte injury. Additionally, among the down-regulated genes, those with prior reports as podocyte injury markers (FMC1 and LAMC3) were selected as candidates for validation.

DEG analysis between subgroups (SSNS vs. SRNS) identified 2,350 genes to be significant (P<0.5), with 1,409 up-regulated and 941 down-regulated genes. Heatmap showed that the gene expression pattern varied as per the disease subgroup (Fig. 3). To identify a significant gene signature contributing to variability in gene expression between the SSNS and SRNS, a fold change cutoff of -1< log FC >1 was applied. This filtering identified a total of 46 genes, 24 up-regulated and 22 down-regulated genes in SSNS compared to SRNS that are listed to the right of heatmap (Fig. 3). GO analysis revealed these genes code for enzymes involved in transcription or translation regulation, cell cycle regulation, apoptotic pathways, membrane trafficking, and signal transduction (Supplementary Table 6). Pathway analysis of these genes did not reveal involvement of any significant pathway due to small number of genes. However, among the significant genes, FMC1, LTBR, TAGLN2, and BAP1 had earlier reports of being involved in nephrotic syndrome and were thus selected for quantitative real time-polymerase chain reaction (qRT-PCR) validation.

The expression pattern of selected candidate genes (ID1, ID3, FOSB, SOCS3, SMAD7, FMC1, and LAMC3), was similar in qRT-PCR and microarray (correlation of 0.95 and 0.89 of SSNS and SRNS respectively) (Fig. 4A, B). qRT-PCR also confirmed the expression pattern of candidate genes (FMC1, LTBR, BAP1, and TAGLN2) between the SSNS and SRNS with a correlation of 0.71 (Fig. 4C).