Gene Expression as a Guide to the Development of Novel Therapies in Primary Glomerular Diseases

Despite improvements in understanding the pathogenic mechanisms of primary glomerular diseases, therapy still remains nonspecific. We sought to identify novel therapies targeting kidney-intrinsic injury of distinct primary glomerulonephritides through computational systems biology approaches. We defined the unique transcriptional landscape within kidneys from patients with focal segmental glomerulosclerosis (FSGS), minimal change disease (MCD), immunoglobulin A nephropathy (IgAN), membranous nephropathy (MN) and thin basement membrane nephropathy (TBMN). Differentially expressed genes were functionally annotated with enrichment analysis, and distinct biological processes and pathways implicated in each primary glomerular disease were uncovered. Finally, we identified novel drugs and small-molecule compounds that may reverse each glomerulonephritis phenotype, suggesting they should be further tested as precise therapy in primary glomerular diseases.


Introduction
Primary glomerulonephritides encompass a heterogeneous group of glomerular diseases characterized by abnormal activation of innate and/or adaptive immune responses due to kidney-intrinsic factors [1]. Although relatively rare, they represent the most common cause of end-stage renal disease in young adults and are associated with increased morbidity, mortality and healthcare costs [2]. Current clinicopathological classification includes minimal change disease (MCD), focal segmental glomerulosclerosis (FSGS), membranous nephropathy (MN), immunoglobulin and complement-mediated glomerular diseases with a membranoproliferative glomerular pattern (MPGN), immunoglobulin A nephropathy (IgAN) and thin basement membrane nephropathy (TBMN) [1]. Despite improvements in understanding the underlying pathogenic mechanism of each glomerular disease, therapy still remains nonspecific and includes general supportive measures coupled with immunosuppression [3,4]. Disease-specific therapy targeting kidney-intrinsic injury still remains a major challenge in nephrology.
Recent advances in omics technologies have provided insights into the molecular mechanisms underlying complex traits, such as glomerular diseases. Gene expression represents the intermediate phenotype between genetic variation and disease phenotypic variation and thus may inform about genetic and environmental effects on cells and tissues.
Specifically, the comparison of gene expression variation between distinct conditions can delineate transcriptional differences and specific molecular pathways [5]. To this end, highthroughput genome-wide gene expression studies have described the transcriptome of the peripheral blood and kidneys of animal models and patients with glomerular diseases and uncovered molecular pathways implicated in their pathogenesis [6]. However, gene expression patterns unique to each primary glomerular disease remain to be defined.
Computational systems biology combines knowledge-driven experimental data with simulation-based analyses and tests hypotheses with in silico experiments. This provides a powerful tool to understand complex biological processes and identify novel drugs or drugs to be repurposed [7]. The Connectivity Map (CMap) project was the first approach that provided genome-wide gene expression responses from 4 human cell lines treated with 1 out of 1309 FDA-approved drugs or small-molecule compounds at different doses [8]. The NIH Library of Integrated Network-based Cell-Signatures (LINCS) program expanded the CMap project to include over a million signatures through the use of the L1000 high-throughput transcriptomic technology; it identified gene expression alterations (before and after treatment) of more than 60 human cell lines with more than 20,000 drugs/small-molecule compounds [9,10]. By the use of a multivariate method to compute signatures, the LINCS L1000 Characteristic Direction Signatures Search engine (L1000CDS 2 ) further prioritized thousands of small-molecule signatures and their pairwise combinations, predicted to either mimic or reverse the gene expression signature of a disease or condition [11]. Recently, connectivity mapping identified BI-2536 as a potential drug to treat diabetic nephropathy [12].
Herein, we employed the Nephroseq classic v4 web-based analysis engine [13], a platform for integrative data mining of comprehensive kidney disease gene expression datasets, and identified datasets that compared kidney gene expression patterns from patients with primary glomerular diseases vs. healthy individuals. We defined the unique transcriptional landscape of FSGS, MCD, IgAN, MN and TBMN through the comparison of kidney gene expression patterns between the yielded datasets. Differentially expressed genes were functionally annotated with enrichment analysis, and distinct biological processes and pathways implicated in each primary glomerular disease were established. Finally, using the L1000CDS 2 engine [11], we identified putative novel drugs or small-molecule compounds that may reverse each glomerulonephritis phenotype, suggesting they should be further tested as precise therapy in primary glomerular diseases.

Search Strategy
To identify publicly available gene expression profiles of human glomeruli from patients with primary glomerular diseases, we used the Nephroseq classic v4 web-based engine [13]. Gene expression datasets generated using the Affymetrix Microarray Technology were only included in the study. Specifically, the keywords "human", "kidney", "glomerulus", "microarray" and "affymetrix" were used in the engine to search for gene expression patterns of diseases. Expression profiles from minimal change disease (MCD), focal segmental glomerulosclerosis (FSGS), membranous nephropathy (MN), immunoglobulin A nephropathy (IgAN) and thin basement membrane nephropathy (TBMN), when compared to expression profiles from healthy kidneys, were yielded. We applied a p-value cutoff of 0.05, as provided by the Nephroseq classic v4 web-based engine, to define the statistically significant differentially expressed genes (DEGs). All DEGs that demonstrated statistical significance, irrespective of their fold-change, were included in the analysis. Comparison between DEGs was performed with Venny 2.1 [14] and InteractiVenn [15].

Functional Enrichment Analysis and Drug Prediction
Enrichment analysis of DEGs was performed using g:Profiler [16]. The visualization of enrichment analysis was carried out using the GraphPad Prism version 5.0 [17]. To identify compounds that efficiently reverse the disease-specific transcriptional signa-tures, the L1000CDS 2 engine was used [11]. To identify the possible mechanism of action, disease target, side effects and FDA approval of the drugs/small molecules, we used the Large-Scale Visualization of Drug-induced Transcriptomic signatures (L1000FWD) engine [18], the DrugBank resource [19] and the FDA website.

Computational Systems Biology Approaches Reveal Glomerulonephritis-Specific Gene Signatures
We initially employed the Nephroseq classic v4 web-based analysis engine [13] to identify datasets of kidney gene expression patterns from patients with primary glomerular diseases and healthy individuals. Therefore, we collected previously published datasets that compared kidney gene expression from patients with focal segmental glomerulosclerosis (FSGS) (a total of 39 patients) [20,21], minimal change disease (MCD) (a total of 21 patients) [20,21], IgA nephropathy (IgAN) (a total of 54 patients) [21,22], membranous nephropathy (MN) (21 patients) [21] or thin basement membrane nephropathy (TBMN) (3 patients) [21], with a total of 36 healthy individuals. Clinical parameters and demographics of patients and healthy individuals are presented in Table S1.
To define the unique transcriptional landscape of each primary glomerular disease, we compared gene expression patterns between the yielded datasets. Differentially expressed genes (DEGs) that were exclusively identified in one glomerular disease (and not in others) defined the glomerulonephritis-specific gene signature. Accordingly, 202, 150, 671, 2058 and 1233 DEGs defined the FSGS-specific, MCD-specific, IgAN-specific, MN-specific and TBMN-specific gene signature, respectively, suggesting a unique pathogenic involvement highly specific to the respective glomerular disease. Interestingly, no common DEGs across all diseases were identified ( Figure 1, Table S2). These data demonstrate that gene expression patterns within kidneys in primary glomerular diseases are deregulated in a glomerulonephritis-specific manner. Experimental validation of differential transcripts and their functionality in kidneys is further needed.

Functional Enrichment Analysis and Drug Prediction
Enrichment analysis of DEGs was performed using g:Profiler [16]. The visualization of enrichment analysis was carried out using the GraphPad Prism version 5.0 [17]. To identify compounds that efficiently reverse the disease-specific transcriptional signatures, the L1000CDS 2 engine was used [11]. To identify the possible mechanism of action, disease target, side effects and FDA approval of the drugs/small molecules, we used the Large-Scale Visualization of Drug-induced Transcriptomic signatures (L1000FWD) engine [18], the DrugBank resource [19] and the FDA website.

Computational Systems Biology Approaches Reveal Glomerulonephritis-Specific Gene Signatures
We initially employed the Nephroseq classic v4 web-based analysis engine [13] to identify datasets of kidney gene expression patterns from patients with primary glomerular diseases and healthy individuals. Therefore, we collected previously published datasets that compared kidney gene expression from patients with focal segmental glomerulosclerosis (FSGS) (a total of 39 patients) [20,21], minimal change disease (MCD) (a total of 21 patients) [20,21], IgA nephropathy (IgAN) (a total of 54 patients) [21,22], membranous nephropathy (MN) (21 patients) [21] or thin basement membrane nephropathy (TBMN) (3 patients) [21], with a total of 36 healthy individuals. Clinical parameters and demographics of patients and healthy individuals are presented in Table S1.
To define the unique transcriptional landscape of each primary glomerular disease, we compared gene expression patterns between the yielded datasets. Differentially expressed genes (DEGs) that were exclusively identified in one glomerular disease (and not in others) defined the glomerulonephritis-specific gene signature. Accordingly, 202, 150, 671, 2058 and 1233 DEGs defined the FSGS-specific, MCD-specific, IgAN-specific, MNspecific and TBMN-specific gene signature, respectively, suggesting a unique pathogenic involvement highly specific to the respective glomerular disease. Interestingly, no common DEGs across all diseases were identified ( Figure 1, Table S2). These data demonstrate that gene expression patterns within kidneys in primary glomerular diseases are deregulated in a glomerulonephritis-specific manner. Experimental validation of differential transcripts and their functionality in kidneys is further needed.

Unique Biological Processes and Pathways Are Implicated in Distinct Primary Glomerular Diseases
Next, to reveal altered molecular pathways between primary glomerular diseases, glomerulonephritis-specific DEGs were functionally interpreted through enrichment analysis using g:Profiler [16]. To this end, the FSGS-specific gene signature was functionally enriched in gene ontology (GO) biological processes linked to cell-cell junction assembly and the enzyme-linked receptor protein signaling pathway. Similarly, the MCD-specific gene signature was functionally enriched in glycine serine threonine metabolism; the IgANspecific gene signature was enriched in regulation of postsynaptic membrane potential and the cAMP signaling pathway; the MN-specific gene signature was enriched in IL-17 and MAPK signaling pathways, whereas the TBMN-specific gene signature was enriched in pathways linked to B-cell receptor signaling and regulation of actin cytoskeleton ( Figure 2). Collectively, these data demonstrate the involvement of unique biological processes and pathways in distinct glomerular diseases, suggesting that these processes and pathways may represent disease-specific novel therapeutic targets. To identify putative novel drugs or small-molecule compounds that may reverse each glomerulonephritis phenotype, we used the L1000 Characteristic Direction Signature Search Engine (L1000CDS 2 ) [16]. The L1000CDS 2 was queried with upregulated and downregulated genes from each glomerular disease to prioritize matching signatures obtained from more than 20,000 drug and small-molecule treatments of multiple human cell lines. Queries generated predictions for the top 50 molecules that could reverse the input of each glomerular disease (Table S3). To identify putative novel drugs or small-molecule compounds that may reverse each glomerulonephritis phenotype, we used the L1000 Characteristic Direction Signature Search Engine (L1000CDS 2 ) [16]. The L1000CDS 2 was queried with upregulated and downregulated genes from each glomerular disease to prioritize matching signatures obtained from more than 20,000 drug and small-molecule treatments of multiple human cell lines. Queries generated predictions for the top 50 molecules that could reverse the input of each glomerular disease (Table S3).
To identify glomerulonephritis-specific novel drugs or small-molecule compounds, we compared the top 50 drug signatures between the 5 glomerular diseases. Drugs or smallmolecule compounds that were predicted to reverse only one glomerulonephritis-specific gene signature and not others defined the glomerulonephritis-specific drug signature. Thus, 17, 9, 13, 8 and 27 drugs/small-molecule compounds were predicted to reverse the FSGS-specific, MCD-specific, IgAN-specific, MN-specific and TBMN-specific gene signature, respectively ( Figure 3, Table 1).
in. Med. 2021, 10, x FOR PEER REVIEW To identify glomerulonephritis-specific novel drugs or small-molecule c we compared the top 50 drug signatures between the 5 glomerular disease small-molecule compounds that were predicted to reverse only one glomeru specific gene signature and not others defined the glomerulonephritis-specif nature. Thus, 17, 9, 13, 8 and 27 drugs/small-molecule compounds were pred verse the FSGS-specific, MCD-specific, IgAN-specific, MN-specific and TBM gene signature, respectively ( Figure 3, Table 1).    Interestingly, PD-0325901 (an investigational MEK inhibitor [23]), PD-184352 (an investigational MAPK inhibitor [24]) and KU 0060648 trihydrochloride were predicted to reverse gene expression patterns of all glomerular diseases ( Table 2). Finally, the potential mechanism of action of identified drugs and small molecules, the disease they may target, their possible side effects and their possible FDA approval were identified through the L1000FWD engine [18], the DrugBank resource [19] and the FDA website (Table 3). Table 3. Potential mechanism of action, target disease, side effects and FDA approval of drugs and small molecules. MOA: mechanism of action; N/A: not available.   Collectively, our data reveal novel, not previously identified, drugs and small-molecule compounds that may reverse the phenotype of primary glomerular diseases in a diseasespecific manner. Further investigation is called for to define the therapeutic implications of our findings.

Discussion
Herein, we employed the Nephroseq classic v4 web-based analysis engine and identified gene expression datasets within kidneys from patients with primary glomerular diseases vs. healthy individuals. The comparison of gene expression patterns defined the FSGS-specific, MCD-specific, IgAN-specific, MN-specific and TBMN-specific gene signatures. Functional enrichment analysis of differentially expressed genes identified distinct biological processes and pathways implicated in each primary glomerular disease. Finally, by the use of computational systems biology resources, we revealed novel, not previously identified, drugs and small-molecule compounds that may reverse the phenotype of primary glomerular diseases in a glomerulonephritis-specific manner.
We elected to use the Nephroseq classic v4 web-based analysis engine that gives access to genome-wide gene expression datasets generated by renal researchers [13]. We yielded datasets generated from hybridization-based analyses of human kidneys from patients with primary glomerular diseases when compared to healthy kidneys. Specifically, we obtained datasets for MCD, FSGS, MN, IgAN and TBMN. Although these diseases are characterized by distinct pathogenic mechanisms and express distinct manifestations, therapy is common and includes general measures to reduce proteinuria and manage hypertension, edema, hyperlipidemia and hypercoagulability, coupled with immunosuppression, such as steroids, alkylating agents, mycophenolate mofetil, calcineurin inhibitors and others) [3,4,25]. To reveal glomerulonephritis-specific processes and pathways implicated in each glomerular disease that could serve as glomerulonephritis-specific targets, we compared the kidney transcriptome between the yielded primary glomerular diseases. We uncovered DEGs exclusively expressed in each glomerular disease that defined each glomerulonephritis-specific gene signature. Gene signatures were further interpreted with functional enrichment analysis, by which delineated processes and pathways unique to each glomerular disease were uncovered.
FSGS and MCD are primary podocytopathies that underlie idiopathic nephrotic syndrome. FSGS is a heterogeneous disease that represents a common phenotypic expression of different clinical and histological syndromes caused by distinct causes, including mutations in podocyte components, circulating factors, viral infections, toxicities and maladaptive responses [26][27][28]. Specifically, the pathogenesis involves a factor disturbing podocyte function and permeability and a T-cell abnormality leading to cytokine production [29,30]. Cardiotrophin-like cytokine-1 (CLC-1) and soluble urokinase plasminogen activator surface receptor (suPAR) are implicated in the pathogenesis of FSGS [31,32]. MCD represents the most common cause of nephritic syndrome in children. Underlying causes may include Hodgkin disease, atopy, exposure to allergens and nonsteroidal anti-inflammatory drugs [33][34][35]. In MCD, deregulated T-cell function can be attributed to altered expression of NF-κB and NFRKB [36,37]. Our data demonstrate that gene expression deregulation within kidneys of human FSGS is associated with biological processes linked to cell-cell junction assembly and the enzyme-linked receptor protein signaling pathway. Gene expression deregulation within kidneys from human MCD is linked to glycine, serine and threonine metabolism. These suggest that the identified processes and pathways may represent targeted therapies for FSGS or MCD.
The current model of IgAN pathogenesis involves a multiple-hit hypothesis model. Genetic and environmental factors predispose individuals to abnormal immune responses to pathogens. Mucosal-derived pathogen-associated molecular patterns induce polyclonal lymphocyte proliferation in a TLR-mediated manner. IgA-secreting plasma cells migrate to mucosa, releasing dimeric IgA into the lumen, whereas misdirected cells are released to systemic circulation, secreting poorly O-galactosylated IgA1. Circulating galactose-deficient IgA1 forms immune complexes with specific antibodies, resulting in immune-complex deposition or in situ formation in the mesangium, inducing mesangial cell proliferation and increased production of mesangial matrix [38,39]. Despite improvements in understanding the pathogenesis of IgAN, this multiple-hit hypothesis model does not explain kidney injury associated with mesangial IgA deposition. Our analysis demonstrates that within kidneys of patients with IgAN, gene expression deregulation is associated with neuroactive ligand-receptor interaction and the cAMP signaling pathways, suggesting they could represent targets of kidney injury in IgAN.
MN is an autoimmune disease characterized by the formation of immune deposits and complement-mediated proteinuria [40]. Epitopes and antigens, such as phospholipase A2 receptor (PLA2R) [41], thrombospondin domain-containing 7A (THSD7A) [42] and neutral endopeptidase [43], have been identified. The role of complement activation mediating podocyte injury in MN is well established [44,45]. Recently, the altered T H 17/T reg ratio was suggested as a mechanism involved in the pathogenesis of idiopathic MN [46]. In agreement, we reveal that within kidneys of patients with MN, the transcriptome deregulation is functionally enriched in IL17 signaling pathways. We also uncover the role of CLEC7A (dectin) and MAPK signaling pathway, suggesting their involvement in tissue injury in idiopathic MN.
TBMN refers to familial and sporadic isolated hematuria associated with an attenuated glomerular basement membrane. It is an autosomal dominant condition caused by mutations in COL4A3 or COL4A4 gene [47]. In this study, we demonstrate that within kidneys from patients with TBMN, DEGs are functionally enriched in pathways of B-cell receptor signaling and regulation of actin cytoskeleton.
Next, we used the identified glomerulonephritis-specific gene signatures, and through computational systems biology approaches, we sought to discover novel drugs or drugs to be repurposed in a disease-specific manner. Specifically, we used the L1000CDS 2 engine [11] and prioritized the top 50 molecules predicted to reverse upregulated and downregulated DEGs of each glomerular disease. Then, drug signatures of the five glomerulonephritides were compared, and glomerulonephritis-specific drugs were identified. Among other drugs, the third-generation kinase inhibitor indicated for the treatment of chronic myeloid leukemia, ponatinib [48], and the JAK3 kinase inhibitor IV [49] were predicted to specifically reverse the FSGS-specific gene signature; pracinostat, an oral histone deacetylase inhibitor [50] was predicted to specifically reverse the MCD-specific gene signature; BMS-754807, an inhibitor of insulin-like growth factor-1R/IR [51], was predicted to specifically reverse the IgAN-specific gene signature; TCPA-1, a dual inhibitor of STAT3 and NF-κB [52], was predicted to specifically reverse the MN-specific gene signature; and wortmannin, an autophagy inhibitor [53], was predicted to specifically reverse the TBMNspecific gene signature. This computational analysis prioritized novel drugs and drugs to be repurposed as glomerulonephritis-specific agents. Further investigation is ongoing to define the therapeutic implications of our findings.
Collectively, through computational systems biology approaches, we defined the FSGS-specific, MCD-specific, IgAN-specific, MN-specific and TBMN-specific gene signatures within human kidneys and identified distinct biological processes and pathways implicated in each disease. We also revealed novel, not previously identified, drugs and small-molecule compounds that may reverse the phenotype of primary glomerular diseases in a glomerulonephritis-specific manner. Experimental validation of the differential transcripts and their functionality is further needed. To this end, future studies will define the pathogenic and therapeutic implications of our findings.

Supplementary Materials:
The following are available online at https://www.mdpi.com/article/ 10.3390/jcm10112262/s1, Table S1: Clinical parameters and demographics of patients and healthy individuals; Table S2: Glomerulonephritis-specific differentially expressed genes; Table S3: Drugs and small-molecule compounds predicted to reverse each primary glomerular disease.