Background

Chronic kidney disease (CKD) represents a significant global public health concern, contributing to increased risks of cardiovascular disease, hospitalization, and mortality. Hemodialysis is a common treatment for CKD, particularly in the advanced stages of the condition. However, its limitations in eliminating all uremic toxins necessitate a deeper understanding of the pathological mechanisms activated by non-dialyzable uremic toxins and the identification of potential therapeutic interventions [29, 45, 47, 55, 69, 92].

Among the key cellular players involved in toxin filtration are renal tubular cells, alongside glomerular cells. Notably, renal tubular cells are at the forefront of exposure to uremic toxins, making them crucial targets for investigation due to their vital role in kidney function and susceptibility to toxin-induced injury [29, 45, 47, 69, 92].

Uremic toxins fall into three categories: small water-soluble solutes, medium molecules, and protein-bound solutes [31]. While numerous uremic toxins contribute to chronic kidney disease (CKD) progression, prior research has primarily focused on indoxyl sulfate (IS) and p-cresyl sulfate (PCS), two protein-bound compounds. These studies unveiled that IS activates the aryl hydrocarbon receptor (Ahr), a ligand-activated transcription factor receptor. Research indicates that Ahr activation correlates with vascular inflammation, leukocyte activation, thrombosis, reactive oxygen species, and cardiotoxicity [70]. Additionally, IS triggers the activation of the epidermal growth factor receptor (EGFR), thereby promoting renal tissue remodeling and arteriosclerosis [80]. IS expedites kidney fibrosis by upregulating expressions of transforming growth factor beta (TGF-β), tissue inhibitors of metalloproteinases-1 (TIMP-1), and pro-collagen [57]. Within kidney tubular cells, IS accumulation wreaks havoc on the anti-oxidative system, fostering cellular dysfunction and heightened oxidative stress [26]. Moreover, IS amplifies the expression of plasminogen activator inhibitor-1 (PAI-1), leading to renal tubular cell dysfunction [58].

PCS exhibits renal toxicity akin to IS [31]. Investigations have unveiled that PCS triggers cellular immune and inflammatory reactions, notably activating the TGF-β signaling pathway [79]. Furthermore, PCS activates rat sarcoma (RAS) and augments oxidative stress generation by stimulating leukocytes, thereby instigating renal tubular epithelial-to-mesenchymal transition. These alterations significantly fuel the advancement of kidney fibrosis [78].

Prior studies have investigated the impact of individual uremic toxins, such as IS and PCS, protein-bound compounds, on renal tubular cells, revealing their tubule-toxic effects and contributions to CKD progression. However, a comprehensive understanding of the effects of non-dialyzable uremic toxins, which can be of any type, on renal tubular cells remains lacking, emphasizing the need for further research to unravel their intricate responses.

This study aims to adopt a systems biology approach to gain a holistic understanding of the comprehensive pathological mechanisms activated by non-dialyzable toxins and identify essential proteins as potential targets for therapeutic intervention.

The primary objectives of this study are twofold: (1) to gain insights into the comprehensive pathological mechanisms of retained uremic toxins on healthy tubular cells, and (2) to identify essential proteins as potential targets for antagonist drugs to control CKD progression. Focusing on antagonist drugs emerges as a justifiable approach for treating CKD due to their proven efficacy in slowing disease progression [59, 89].

Methodology

A flowchart illustrating the overall methodology and data analysis pipeline is provided in Fig. 1. The figure summarizes the step-by-step approach, from data acquisition to pathway enrichment analysis, employed in this study. These steps are explained in the following subsections.

Fig. 1
figure 1

Flowchart of the study’s methodology and data analysis pipeline

Full size image

Data acquisition

Transcriptomics data for this study were obtained from the GEO database under the accession number GSE45709 (n.d.). The dataset includes gene expression measurements from healthy renal tubular cells exposed to both healthy plasma (control group) and post-dialysis uremic plasma (case group). All samples from each group were utilized to ensure consistency and statistical power in the analysis.

By utilizing this dataset, we were able to investigate the differential gene expression profiles between the two groups and identify potential cellular mechanisms underlying the pathological effects of non-dialyzable uremic toxins in CKD.

Data normalization

This study seeks to investigate the effects of non-dialyzable uremic toxins on the gene expression of healthy renal tubular cells. In other words, we seek to measure the variance of gene expression between case (post-dialysis plasma) and control (healthy plasma) groups. However, in addition to this variance, there are also technical and biological variances between samples. To measure the intended variance, we must neutralize the others [13]. Biological variance can be neutralized by repeating samples and technical variance by normalization. We have several samples in each group, which neutralizes the biological variance. For technical variance, we used quantile normalization. This method assumes that technical variance appears as differences in the general characteristics of samples. So, quantile normalization equalizes the statistical distribution of gene expression values across samples, effectively removing technical variance [36]. Figure 2 shows the distribution of gene expression values of samples before and after utilizing quantile normalization.

Fig. 2
figure 2

a and b Show the distribution of log of gene expression values of samples before and after quantile normalization

Full size image

Identification of differentially expressed genes (DEGs)

To identify DEGs between the control and case groups, we employed the limma package in R programming language, a widely accepted method for microarray data analysis that provides reliable results [67], to perform the statistical t-test. We used the FDR technique for adjusting the p-value and genes with a p-value less than 0.05 were considered differentially expressed. We used the fold change (FC) metric to determine the upregulation (log FC > 0) or downregulation (log FC < 0) of DEGs. As the focus of this study is on potential antagonist drug targets, we specifically examined up-regulated genes.

Construction of protein–protein interaction (PPI) network

To explore potential interactions between the up-regulated DEGs, we constructed a PPI network using the STRING server. The network was built using active interaction sources, including text mining, experiments, and databases, with a minimum required interaction score of 0.7 (high confidence).

Identification of essential proteins

Previous studies have shown that centrality measures are capable of identifying essential proteins in PPI networks [19, 28, 98]. So, to identify essential proteins in the PPI network, we assessed several centrality measures, including closeness, betweenness, degree, and eigenvector centrality. Specific thresholds were applied for each centrality measure to identify essential proteins. The thresholds were 8E-4 for closeness, 1E-9 for betweenness, 3 for degree, and 1.24E-6 for eigenvector, based on prior knowledge.

Clustering of essential proteins

We extracted the PPI network of essential proteins from the STRING server using active interaction sources and a minimum required interaction score of 0.4. To identify potential functional modules, we employed a heuristic method based on modularity optimization proposed by Blondel et al. [6]. This approach maximizes the modularity of the resulting clusters, revealing densely connected groups of proteins.

Pathway enrichment analysis

To gain insights into the functions of each protein cluster, we performed pathway enrichment analysis using the Enrichr server. The WikiPathway 2021 Human Library was utilized to identify enriched biological pathways associated with each cluster. Pathways were ranked by adjusted p-value [12].

Results

In this section, we present the results of our data analysis pipeline, aiming to gain insights into the effects of non-dialyzable uremic toxins on healthy renal tubular cells and identify potential therapeutic targets.

Identifying essential proteins

After performing the t-test, we identified 1503 DEGs, with 983 up-regulated and 520 down-regulated genes. Among the up-regulated DEGs, a PPI network analysis using centrality measures (degree, betweenness, closeness, and eigenvector centrality) and predefined thresholds allowed us to identify 70 essential proteins. The resulting PPI network of the 70 essential proteins is depicted in Fig. 3, comprising 195 edges representing protein–protein interactions. In this figure, the size of the node represents the degree. The larger the size, the higher the degree of the node. The color of the node represents the betweenness. The color of the node changes in a spectrum from red to blue. Red indicates a higher betweenness, and blue indicates a lower betweenness.

Fig. 3
figure 3

PPI network of essential proteins

Full size image

Modularity-based clustering

The PPI network of essential proteins was subjected to modularity-based clustering, leading to the identification of six distinct clusters of essential proteins which are depicted in Fig. 4. The first cluster, cluster 0, comprises 19 interactions and 13 proteins, including MYOD1, CDH15, CSF3R, H2AFJ, DHH, GATA4, WNT1, FBXO32, MYH4, WNT11, ZFPM1, PBX1, and LMX1B.

Fig. 4
figure 4

Modularity classes of essential proteins

Full size image

The second cluster, cluster 1, encompasses 38 interactions and 17 proteins, including ABCB8, ATP12A, ABCC8, FXYD1, GCK, INS, PRKACG, AGT, OXT, RHO, GHRL, AVPR2, AGTR1, GNG8, GPRASP1, CACNA2D4, and KISS1.

The third cluster, cluster 2, encompasses 15 interactions and nine proteins, including GFAP, CXCL12, VWF, RGS16, DCX, SYP, S100B, SYN1, and SLC18A2.

The fourth cluster, cluster 3, encompasses 37 interactions and 12 proteins, including KCNB1, KCND3, CAV3, CACNA2D1, HCN4, SCN2B, SCN4A, KCNE1L, KCNA3, KCNA1, KCNV1, and KCNH4.

The fifth cluster, cluster 4, comprises ten interactions and five proteins, including COX4I2, COX6B1, UQCRFS1, NDUFA4L2, and COX7C.

The sixth cluster, cluster 5, comprises 31 interactions and 13 proteins: CHMP2B, RPS2&A, FGF22, FAU, RPS21, ISG15, RPS29, RPS10, FGF17, FGF4, INSRR, FLRT3, and PSMB8. Notably, these clusters represent functionally related groups of essential proteins, potentially serving distinct roles in response to retained uremic toxins.

Enrichment analysis

After clustering, we performed enrichment analysis to identify biological pathways activated by proteins of each cluster and map proteins to their functions. We considered pathways with adjusted p-value < 0.01 as statistically significant pathways. The important enriched pathways associated with each cluster are summarized in the following tables:

  • Cluster 0: Significantly enriched pathways are listed in Table 1. These pathways had adjusted p-value < 0.01.

  • Cluster 1: Significantly enriched pathways are listed in Table 2. These pathways had adjusted p-value < 0.01.

  • Cluster 2: Significantly enriched pathways are presented in Table 3. These pathways had adjusted p-value < 0.01.

  • Cluster 3: We found no statistically significant pathways for proteins of cluster 3.

  • Cluster 4: Significantly enriched pathways are listed in Table 4. These pathways had adjusted p-value < 0.01.

  • Cluster 5: Significantly enriched pathways are presented in Table 5. These pathways had adjusted p-value < 0.01.

Table 1 Enriched pathways of cluster 0
Full size table
Table 2 Enriched pathways of cluster 1
Full size table
Table 3 Enriched pathways of cluster 2
Full size table
Table 4 Enriched pathways of cluster 4
Full size table
Table 5 Enriched pathways of cluster 5
Full size table

Discussion

In this study, we aimed to gain a holistic view of the pathological effects of non-dialyzable uremic toxins on healthy renal tubular cells by adopting a systems biology approach. Our analysis revealed 983 up-regulated DEGs and 70 essential proteins. Subsequently, we identified six dense communities of essential proteins, each representing potential functional modules within the network.

Through pathway enrichment analysis, we identified 24 statistically significant pathways associated with the identified clusters. While these pathways were not directly associated with renal tubular dysfunction, we found indirect evidence linking them to cellular processes involved in inflammation, fibrosis, apoptosis, metabolic dysfunction, and oxidative stress. This association is presented in Table 6. This table shows the participation of each pathway in the mentioned pathological processes.

Table 6 Mapping pathological pathways to their roles in renal tubular dysfunction
Full size table

Among the 24 pathways, 11 were found to be involved in these cellular processes, suggesting their potential relevance to renal tubular dysfunction. To further explore the role of essential proteins in the identified pathways, we investigated their participation rates. To this end, we investigated each pathway. Notably, 22 essential proteins, including WNT1/11, AGT, FGF4/17/22, LMX1B, GATA4, CXCL12, KISS1, COX6B1/7C, UQCRFS1, AGTR1, NDUFA4L2, INS, RGS16, RPS10/21/27A, FAU, and SLC18A2 were found to play a role in the pathways associated with inflammation, fibrosis, apoptosis, metabolic dysfunction, and oxidative stress. Figure 5 shows the participation of these proteins in the 11 identified pathological pathways. WNT1 participates in 4 pathological pathways. WNT 11 participates in 3 pathological pathways. Each of AGT and FGF4/7/22 participates in 2 pathological pathways. Each of LMX1B, GATA4, CXCL12, KISS1, COX6B1, UQCRFS1, AGTR1, NDUFA4LC, COX7C, INS, RGS16, RPS27A, FAU, RPS21, RPS10 and SLC18A2 participates in 1 pathological pathway.

Fig. 5
figure 5

The participation rate of essential proteins in pathological pathways which lead to renal tubular dysfunction. WNT1 participates in 4 pathological pathways. WNT 11 participates in 3 pathological pathways. Each of AGT and FGF4/7/22 participates in 2 pathological pathways. Each of LMX1B, GATA4, CXCL12, KISS1, COX6B1, UQCRFS1, AGTR1, NDUFA4LC, COX7C, INS, RGS16, RPS27A, FAU, RPS21, RPS10 and SLC18A2 participates in 1 pathological pathway

Full size image

WNT1/11

Figure 5 shows that WNT1 and WNT11 have the highest participation rate in pathological pathways of renal tubular cells. WNT1 participates in breast cancer pathway WP4262, dopaminergic neurogenesis WP2855, ESC pluripotency pathways WP3931, and Wnt signaling in kidney disease WP4150. WNT11 participates in breast cancer pathway WP4262, ESC pluripotency pathways WP3931, and Wnt signaling in kidney disease WP4150.

WNT1/11 plays a pivotal role in kidney development and function, modulating proliferation, differentiation, and apoptosis [66]. However, its upregulation leads to renal cell dysfunction. Previous studies have reported elevated levels of WNT1 and WNT11 in various kidney diseases, including cystic kidney diseases, nephropathy, and tubulointerstitial fibrosis [66, 81, 90]. In vivo experiments demonstrated that WNT1/11 overexpression in renal tubules caused dysfunction in tubular cells, supporting their potential role in renal tubular dysfunction [43].

In exploring the potential clinical applications of targeting WNT1/11 to mitigate the pathological pathways activated by non-dialyzable uremic toxins, several promising avenues have emerged. Small molecule inhibitors, exemplified by LGK974 and XAV939, offer a targeted approach to disrupt WNT secretion and downstream signaling components, potentially attenuating renal damage in CKD patients [48, 93]. Monoclonal antibodies like OMP-18R5 and vantictumab could serve as adjunct therapies by specifically blocking WNT proteins, thereby inhibiting aberrant signaling cascades associated with CKD progression [60, 74].

Moreover, RNA interference (RNAi) strategies hold promise in silencing WNT1/11 gene expression, offering a potential avenue for personalized treatment approaches tailored to CKD patients' specific molecular profiles [49, 54]. Additionally, the therapeutic potential of natural compounds such as curcumin, Tripterygium wilfordii, and astragaloside IV cannot be overlooked. Derived from plants or other sources, these compounds exhibit promising capabilities in downregulating WNT1/11 expression or activity through various mechanisms, providing alternative or complementary therapeutic options for CKD management [11, 15, 38, 91].

AGT/AGTR1

Figure 5 shows that AGT/R1 has a high participation rate in pathological pathways of renal tubular cells. AGT participates in ACE inhibitor pathway WP554 and ESC pluripotency pathways WP3931. AGT is produced in the kidney and cleaved by renin to form angiotensin I and II, regulating blood pressure, sodium reabsorption, and fluid balance in renal tubules [96]. However, elevated AGT levels in the kidney have been associated with renal tubular dysfunction in humans and animal models [86, 87]. Studies with AGT overexpression in animal models have observed tubular cell apoptosis, inflammation, fibrosis, and other signs of tubular injury, supporting the association between elevated renal AGT levels and tubular dysfunction [50, 97].

In considering the translation of identified targets into clinical applications, AGT emerges as a promising candidate for therapeutic intervention in CKD. AGT overexpression contributes to renal tubular dysfunction, highlighting its significance as a target for antagonist drugs aimed at mitigating disease progression.

Several pharmacological agents have been proposed to target AGT, primarily for managing conditions like hypertension and heart failure, which commonly coexist with CKD. For example, angiotensin-converting enzyme (ACE) inhibitors such as captopril and enalapril have demonstrated efficacy in reducing AGT-mediated effects on blood pressure and fluid balance, thereby potentially attenuating renal damage [27, 37]. Similarly, angiotensin receptor blockers (ARBs) like losartan and valsartan offer alternative therapeutic avenues by blocking downstream effects of AGT activation, contributing to the regulation of blood pressure and renal function. Additionally, renin inhibitors like aliskiren provide another approach to modulating the renin–angiotensin–aldosterone system, targeting AGT-mediated pathways implicated in CKD progression [17, 73].

FGF4/17/22

As Fig. 5 shows, FGF4/17/22 has a high participation rate in pathological pathways of renal tubular cells. FGF 4/17/22 participate in breast cancer pathway WP4262 and ESC Pluripotency pathways WP3931. FGF signaling plays an important role in kidney development and function by modulating cell proliferation, differentiation, survival, and repair [3, 85]. FGFs are expressed in the adult kidney, particularly in the renal tubules [3, 8]. Some studies have found elevated levels of these FGFs in animal models and patients with kidney diseases like polycystic kidney disease, nephropathy, and tubulointerstitial fibrosis [44, 77, 83].

Expanding on the potential translation of the identified targets into clinical applications, the overexpression of FGF4, FGF17, and FGF22 in renal tubules has emerged as a significant factor contributing to dysfunction. Understanding their context-dependent functions is crucial for developing targeted FGF-based therapies aimed at addressing tubule disorders in CKD patients.

Several therapeutic approaches have been proposed to target FGF4/17/22, offering potential avenues for personalized treatment in CKD. For example, small molecule inhibitors like erdafitinib indirectly inhibit FGF activity by targeting FGFRs, thereby modulating FGF-mediated signaling pathways [56]. Similarly, antibodies such as R3Mab have shown promise in blocking FGF activity by preventing their interaction with receptors or promoting their degradation, offering targeted interventions to mitigate renal dysfunction [35].

RNAi-based approaches, exemplified by molecules like AZD4547, offer a mechanism to silence FGF expression by targeting their mRNA, potentially reducing aberrant FGF signaling and mitigating tubule disorders in CKD [64]. Additionally, peptides like FGF Trap competitively inhibit FGF binding to receptors, disrupting FGF signaling pathways and offering therapeutic potential in CKD management [84].

Furthermore, proteins or molecules like FP-1039/GSK3052230 can bind and sequester FGF ligands, effectively inhibiting FGF signaling pathways and providing a targeted approach to modulating FGF-mediated effects on renal function [5].

LMX1B

LMX1B participates in dopaminergic neurogenesis WP2855. This protein plays an important role in kidney function and development. Previous studies have shown that LMX1B is expressed in renal tubular epithelial cells and is required for proper podocyte differentiation [7]. Our results initially hypothesized that overexpression of LMX1B leads to tubule injury and dysfunction. However, there is also evidence that LMX1B overexpression may have a healing function. Some studies have found that elevated LMX1B expression is seen in regenerating tubules after injury, suggesting that LMX1B may promote renal tubular repair and regeneration [100].

As LMX1B is primarily a transcription factor involved in developmental processes, there are currently no specific molecules designed to target LMX1B directly for therapeutic purposes.

GATA4

GATA4 participates in mammalian disorder of sexual development WP4842. This protein plays an important role in kidney function, particularly in renal tubular cells [14]. Our results hypothesize that GATA4 overexpression leads to renal tubular cell dysfunction and injury. This is supported by evidence from the literature.

Studies have shown that high GATA4 levels in renal tubular epithelial cells correlate with tubular cell dysfunction and injury, contributing to diabetic nephropathy [14]. Both articles find that inhibiting GATA4 activity, such as by promoting its degradation, attenuates tubular cell damage and fibrosis.

Expanding on the potential translation of the identified targets into clinical applications, various strategies have been proposed to target GATA4, offering promising avenues for personalized treatment in CKD.

Compounds like histone deacetylase (HDAC) inhibitors provide indirect modulation of GATA4 activity by affecting proteins interacting with GATA4 or downstream targets. These compounds have shown efficacy in contexts like cardiac regeneration, suggesting their potential utility in mitigating GATA4-related abnormalities associated with CKD [101].

Experimental techniques involving adeno-associated virus (AAV) vectors offer another approach to target GATA4, with the aim of delivering GATA4 gene constructs to cells with mutations, potentially restoring normal expression and function. While initially explored in the context of cardiac disease, similar gene therapy approaches could hold promise for addressing GATA4-related abnormalities in CKD [65].

RNAi approaches, utilizing small interfering RNAs (siRNAs) or antisense oligonucleotides, present a mechanism to silence GATA4 expression at the mRNA level, potentially reducing aberrant protein levels associated with CKD progression [75].

Furthermore, targeting pathways regulated by GATA4, such as Wnt/β-catenin signaling, offers alternative therapeutic approaches. Compounds modulating these pathways, such as Wnt antagonists or β-catenin inhibitors, may offer potential interventions to mitigate GATA4-related abnormalities and improve outcomes for CKD patients [41].

CXCL12

CXCL12 participates in EV release from cardiac cells and their functional effects WP3297. While CXCL12 is known to promote the regeneration of renal tubules after acute kidney injury [71, 95], sustained CXCL12 overexpression may have detrimental effects [76]. Studies involving tubule-specific CXCL12 overexpression in mouse models found that sustained CXCL12 overexpression in adult renal tubular cells led to tubular damage, inflammatory cell infiltration, interstitial fibrosis, and impairment of genes related to electrolyte transport [94].

In exploring the translation of the identified targets into clinical applications, several promising approaches have emerged for targeting CXCL12, offering potential avenues for personalized treatment in CKD.

Small molecule inhibitors like AMD3100 represent one such approach, disrupting the CXCL12-CXCR4 interaction and inhibiting downstream signaling pathways. AMD3100's application in stem cell transplantation and its exploration in cancer therapy highlight its potential utility in mitigating CXCL12-mediated effects in CKD patients [18].

Compounds like NOX-A12 offer another avenue for targeting CXCL12, interfering with its binding to CXCR4. As a PEGylated Spiegelmer, NOX-A12 has been studied in clinical trials for hematological malignancies and solid tumors, suggesting its potential as a therapeutic intervention in CKD-associated conditions [39].

RNAi approaches utilizing siRNAs or antisense oligonucleotides provide yet another mechanism to target CXCL12, aiming to silence its expression at the mRNA level and potentially inhibiting its function in diseases such as cancer and inflammatory disorders [88].

Conclusion and limitations

In conclusion, our systems biology analysis has yielded valuable insights into potential molecular mechanisms contributing to renal tubular dysfunction in CKD. The identification of up-regulated proteins such as WNT1/11, AGT, FGF4/7/17/22, LMX1B, GATA4, CXCL12, KISS1, COX6B1/7C, UQCRFS1, AGTR1, NDUFA4L2, INS, RGS16, RPS10/21/27A, FAU, and SLC18A2 highlights promising targets for further investigation. However, several limitations need to be acknowledged.

Firstly, the sample size of the gene expression data used in this study was very small, comprising only five controls and five cases, which may limit the generalizability of our findings. Additionally, the reliance on a single dataset (GSE45709) may also restrict the generalizability of our findings. Future studies should aim to validate our results using larger and more diverse datasets to enhance the robustness and applicability of our findings.

Moving forward, experimental validation of the identified targets is imperative. Plans or strategies for validation through in vitro cell culture experiments or in vivo animal models based on key target proteins and potential pathways need to be outlined comprehensively. This would provide crucial evidence to support the significance of our findings and facilitate their translation into clinical applications.

Despite these limitations, our study sheds light on the molecular players and pathways associated with renal tubular dysfunction, offering potential avenues for novel therapeutic interventions and enhanced clinical management of CKD and its complications. Emphasizing the need for continued research in this direction will drive progress toward better understanding and tackling CKD, ultimately benefiting patients' outcomes and overall healthcare.