A Differential Transcriptional Regulome Approach to Unpack Cancer Biology: Insights on Renal Cell Carcinoma Subtypes

Abstract

Cancer research calls for new approaches that account for the regulatory complexities of biology. We present, in this study, the differential transcriptional regulome (DIFFREG) approach for the identification and prioritization of key transcriptional regulators and apply it to the case of renal cell carcinoma (RCC) biology. Of note, RCC has a poor prognosis and the biomarker and drug discovery studies to date have tended to focus on gene expression independent from mutations and/or post-translational modifications. DIFFREG focuses on the differential regulation between transcription factors (TFs) and their target genes rather than differential gene expression and integrates transcriptome profiling with the human transcriptional regulatory network to analyze differential gene regulation between healthy and RCC cases. In this study, RNA-seq tissue samples (n = 1020) from the Cancer Genome Atlas (TCGA), including healthy and tumor subjects, were integrated with a comprehensive human TF-gene interactome dataset (1122603 interactions between 1289 TFs and 25177 genes). Comparative analysis of DIFFREG profiles, consisting of perturbed TF-gene interactions, from three common subtypes (clear cell RCC, papillary RCC and chromophobe RCC) revealed subtype-specific alterations, supporting the hypothesis that these signatures in the transcriptional regulome profiles may be considered potential biomarkers that may play an important role in elucidating the molecular mechanisms of RCC development and translating knowledge about the genetic basis of RCC into the clinic. In addition, these indicators may help oncologists make the best decisions for diagnosis and prognosis management.

Introduction

Renal cell carcinoma (RCC) is one of the 10 most common cancers in adults and encompasses >90% of kidney cancers worldwide (Caliskan et al., 2020). It originates from the renal epithelium and is more common in males than in females, and most patients are at an older age. Primarily, RCC is mainly composed of clear cell RCC (ccRCC, also known as KIRC), papillary RCC (pRCC, also known as KIRP), chromophobe RCC (chRCC, also known as KICH), and so on, according to their morphology and immunohistochemical features (Caliskan et al., 2021).

ccRCC is the most common variant, corresponds to 75–80% of all RCCs, and originates from the proximal convoluted tubules of the kidney. The rest are papillary (10–15%), chromophobe (5%), and rare kidney cancers. Although there is improvement of the state-of-the-art treatment technologies, the overall prognosis is still poor in RCCs, because most RCCs are incidentally found at imaging investigations, and about 30% of patients present with metastatic disease at the time of diagnosis (Hsieh et al., 2018). Furthermore, RCC subtypes usually do not respond to conventional radiotherapy and chemotherapy because they have drug resistance. Therefore, focusing on elucidating pathogenesis and biological mechanisms of each subtype of RCC may be promising for early diagnosis, new treatment strategies, and prognosis, and, in turn, might be life-saving.

Gene expression programs are shaped through the interplay of transcription factors (TFs), noncoding RNAs, and their target genes (Cao et al., 2015). Gene regulatory network (GRN) studies reveal complex life events in terms of a descriptive model of such connections introduced by transcriptional and post-transcriptional regulatory datasets, and systems biology approaches integrate heterologous information to understand the function of networks occurring within cells (Emmert-Streib et al., 2014).

In previous years, with the development of high-throughput screening techniques that started with microarray technology and now shifting to next-generation sequencing technologies, a lot of data on gene expressions and their regulations were accumulated and are available to the research community. Various methods and strategies have been developed to create GRNs with a vast amount of information accumulation (Del Vecchio et al., 2017). These regulatory data are used in medicine and molecular biology applications such as identification of disease mechanisms and pathways, the discovery of new drugs and drug repositioning, reducing side-effects of treatments, studying the expression patterns of genes with no evident function, identification of efficacious biomarkers, and development of diagnostic and prognostic screens with high accuracy.

In this study, we investigated the RNA-seq data (n = 1020 samples) form the Cancer Genome Atlas (TCGA) and used the differential regulome methodology that integrates transcriptome data with the human transcriptional regulatory (TF-target gene) network to analyze and compare the differential gene regulations among healthy and tumor groups. Three common subtypes (ccRCC, pRCC, and chRCC) of RCC were investigated and compared in terms of differential regulome profiles. These analyses address the question of which TF-gene interactions are perturbed between the disease and normal conditions and which TFs are likely to change their targets. Differential regulome methodology focuses on the differential regulation between TFs and their target genes rather than differential expression (Fig. 1).

FIG. 1.

Representative illustration of the basis of the DIFFREG methodology. The figure depicts the alteration of TF-gene interactions between the disease and normal phenotype. DIFFREG focuses on the differential regulation between TFs and their target genes rather than differential expression. DIFFREG, differential transcriptional regulome; TF, transcription factor.

Materials and Methods

Data extraction: TF-target gene interactions

To generate a highly reliable TF-target gene interaction dataset, we used several databases that generate data based on experimental values or computational origin, including Regnetwork (Liu et al., 2015), HTRIdb (Bovolenta et al., 2012), IRegulon (Verfaillie et al., 2014), Pazar (Portales-Casamar et al., 2008), TRRUST (Han et al., 2015), and TFacts (Essaghir et al., 2010). The TF-gene interactome set consisted of 1122603 interactions among 1289 TF and 25177 genes.

Regulatory interactions between TFs, microRNAs (miRNAs), and target genes from 25 selected databases for human and mouse are collected, integrated, and reorganized by the Regnetwork database (Liu et al., 2015). Specifically, the human regulatory network contains 23,079 nodes and 369,277 edges, consisting of 1456 TFs, 1904 miRNAs, and 19,719 target genes in the Regnetwork database. The HTRIdb is an open-access database that keeps TF-target gene information with 284 TFs that regulate 18302 genes, totaling 51,871 TF-target gene interactions (Bovolenta et al., 2012).

The Pazar database includes 1433 annotated publications, 1284 regulated genes, 6869 regulatory sequences, and 708 TFs (Portales-Casamar et al., 2008). TRRUST database includes a significantly increased size of the database consisting of 8444 regulatory interactions for 800 TFs in humans (Han et al., 2015). TFacts concludes data of TFs associated with 2720 target genes and 6401 experimentally validated regulations (Essaghir et al., 2010). iRegulon relies on the analysis of regulatory sequences around each gene in the gene set to detect enriched TF motifs or ChIP-seq peaks, using databases of nearly 10.000 TF motifs and 1000 ChIP-seq data sets or “tracks.” Next, it associates enriched motifs and tracks with candidate TFs and determines the optimal subset of direct target genes (Verfaillie et al., 2014).

Identification of differentially expressed genes

RNA-seq tissue samples (n = 1020) from TCGA, including healthy and tumor lineages, were analyzed with the implementation of differential transcriptional regulome (DIFFREG) methodology. RNA-seq count expression values were used for the identification of differentially expressed genes (DEGs), while RNA-seq FPKM-normalized transcriptome datasets were used for the DIFFREG methodology. Among those, 610 samples (72 control and 538 tumors) for ccRCC, 321 samples (32 control and 289 tumors) for pRCC, and 89 samples (24 control and 65 tumors) for chRCC were examined and compared concerning the differential regulome profiles.

Each dataset was statistically analyzed to identify DEGs independently. DEGs were identified from RNA-seq count expression values using the DESeq algorithm (Anders and Huber, 2010), which implements a negative binomial distribution-based method in which mean and variance are linked by local regression to detect DEGs. Analyses were performed in R/Bioconductor, and statistical significance was determined using log2 (fold change) >1 and adjusted p-value criteria (p < 0.05). Benjamini-Hochberg's method was used as a control for the false discovery rate.

Prediction of differential regulome

The DIFFREG method uses three datasets as input: (1) List of DEGs for the corresponding disease, (2) a transcriptome dataset of tissues, including healthy and control samples, and (3) regulatory interaction data (TF-target gene associations). The algorithm first creates a new TF-target gene list containing only DEGs in the target gene column and then performs all subsequent calculations using this new list. DIFFREG calculates the correlation between the expression levels of TFs and target genes in both healthy and diseased states.

It creates groups by taking each TF and its target genes after the correlation calculations. It then performs a significance test (Student's t-test or Mann–Whitney U test) and calculates the p-value over the correlation values for each group between healthy and diseased states. TFs of groups with a p-value <0.05 were identified as key TFs in the disease condition. These gene modules are determined as highlighted groups because the regulation of these target genes changes under different conditions (control vs. diseased). Consequently, DIFFREG estimates the extent to which each TF-gene interaction is perturbed between disease and control cases (Fig. 2). All steps of the algorithm were implemented in the R language.

FIG. 2.

Workflow of the DIFFREG methodology. The algorithm first creates a new TF-target gene list containing only DEGs in the target gene column and then performs all subsequent calculations using this new list. DIFFREG calculates the correlation between the expression levels of TFs and target genes in both healthy and diseased states. It creates groups by taking each TF and its target genes after the correlation calculations. It then performs a significance test (Student's t-test or Mann–Whitney U test) and calculates the p-value over the correlation values for each group between healthy and diseased states. TFs of groups with a p-value <0.05 were identified as key TFs in the disease condition. DEGs, differentially expressed genes.

Pathway enrichment analyses

To in-depth analyze the highlighted TF-target gene modules at the functional level, Kyoto Encyclopedia of Gene and Genome (KEGG) pathway analyses were performed using Database for Annotation, Visualization, and Integrated Discovery (DAVID), which is a web-based online bioinformatics tool that aims to provide a comprehensive pool of functional annotation tools for the researchers to figure out the molecular mechanisms associated with extensive gene lists. p < 0.05 was considered statistically significant (Dennis et al., 2003).

Results

Defining the differential regulome

It is important to distinguish between healthy and afflicted tissues to understand the biological mechanism behind illnesses. The ability to simultaneously consider thousands of genes, thanks to gene expression data, has changed the research of diseases. Therefore, it is not unexpected that a number of recent articles explicitly take gene expression data into account in relation to biomolecular networks.

Although there are mutations or post-translational modifications, which can modify the function of TFs without changing their expression level, the majority of studies on biomarkers and therapeutic targets concentrate on the expression level of genes. This could lead to the omission of TFs that are important in illness. Because of these limitations, it is necessary to shift the focus of disease research from differential gene expression to differential networking. To this end, we applied, in this study, the DIFFREG approach to identify and prioritize the key transcriptional regulators that have a strong impact on the expression of a specific group of genes for each RCC subtype (Fig. 1). The DIFFREG strategy looks for significant TFs, including those that are not DEG, but are extremely important in the onset and development of disease.

Prediction of differential regulome

To gain a deeper understanding of molecular mechanisms that distinguish between subtypes, it is necessary to identify the regulatory patterns that are different for each phenotype. To this end, we designed a method that predicts key TFs and differential regulome networks between healthy and disease conditions. One thousand twenty RNA-seq tissue samples were analyzed using a DIFFREG method that combines transcriptome data with human regulatory data to analyze the different regulations between healthy and diseased phenotypes. Six hundred ten samples (72 control and 538 tumors) for ccRCC, 321 samples (32 control and 289 tumors) for pRCC, and 89 samples (24 control and 65 tumors) for chRCC were analyzed and compared in terms of differential regulome profile. DIFFREG estimates the extent to which each TF-gene interaction is perturbed between disease and control cases (Fig. 2).

Three datasets are used as input for the differential regulome method: (1) Regulatory interaction data, (2) a transcriptome dataset of tissues, (3) and a list of DEGs for the associated disease. Before performing any further calculation, the method first generates a new TF-target gene list with only DEGs in the target gene column. In both healthy and diseased conditions, DIFFREG determines the correlation between the expression levels of TFs and their target genes.

Following the correlation calculations for both healthy and diseased samples, groups are created for each TF by taking that TF and its target genes. Following that, a significance test is run (Student's t-test or Mann–Whitney U test), and the p-value is determined for each group's correlation values between the healthy and diseased states. p-Values of <0.05 were used to identify the important TFs in the disease condition. Because the regulation of these target genes varies depending on the situation, these gene modules were chosen as highlighted groupings. DIFFREG thus calculates the degree of each TF-gene interaction perturbation between disease and control cases (Fig. 2).

Key TFs

DIFFREG algorithm provided the list of TFs whose correlation with its targets was significantly different between tumor and control and their target genes for each subtype (Table 1 and Supplementary File S1). For the ccRCC profile, we identified 820 differential interactions between 48 major TFs and 113 target genes (Fig. 3). The differential regulatory pattern of pRCC showed 795 altered interactions between 31 TFs and 140 target genes (Fig. 3). In chRCC, 404 altered interactions were observed between 37 TFs and 67 target genes (Fig. 3). The major TFs and their target genes were also detected for each subtype considering their cancer specificity (Fig. 3).

FIG. 3.

Key TFs and their target genes were demonstrated for each subtype, also considering their cancer specificity. (A) Representative illustration of the differential regulatory network constructed with significant TFs and their target genes discovered in ccRCC. (B) Depiction of MYB and MYBL2 TFs and their target genes involved in the focal adhesion pathway and steroid hormone synthesis pathway. (C) In ccRCC, the regulation CYP7A1 by AR is depicted. (D) Representative illustration of the differential regulatory network created with significant TFs and their target genes identified in chRCC. (E) Differential regulatory network of significant TFs enriched in maturity-onset diabetes of the young pathway and their target genes discovered in chRCC cases. (F) Representative illustration of TFs and target genes involved in the serotonergic synaptic pathway in chRCC patients. (G) Representative illustration of the differential regulatory network built with significant TFs and their target genes identified in pRCC. AR, androgen receptor; ccRCC, clear cell renal cell carcinoma; chRCC, chromophobe renal cell carcinoma; pRCC, papillary renal cell carcinoma.

Table 1.

Key Transcription Factors Identified by the Differential Transcriptional Regulome Algorithm for Each Subtype

Subtype	Key transcription factors
ccRCC	AFF4, AR, ARNTL, ARX, BARX2, CAT, CCDC25, DDIT3, E2F4, ELF5, ESR1, ETV4, ETV7, FOSB, FOXF2, FOXL1, HOXC8, HSF1, IRF8, ISX, MECOM, MNX1, MYB, MYBL2, NFE2L1, NFE2L2, NKX2-1, NR1I2, PBX2, PHLDA2, PITX2, POLR3A, POU5F1(OCT4), RARB, RAX, RFX1, RXRG, SOX15, SOX8, SSX3, TAL1, TCEAL6, VDR, WT1, ZBTB7A, ZDHHC5, ZMAT2, ZNF503
pRCC	AKR1A1, ANXA11, CEBPB, E2F4, ELF3, ELK4, FOXA3, FOXJ1, GABPB1, GPD1, HINFP, HIST1H2BN, HNF1A, IRF5, IRF7, IRF8, JUND, MXD4, MYCN, MYOG, NFKB1, NFKB2, NFYB, PRNP, RFX4, SOX4, SPI1, SREBF2, TRIM69, USF2, YBX1
chRCC	ALX1, ALX4, ARNT, ARNTL, ARX, BACH2, BARHL2, CDX1, DLX2, DLX4, ESX1, GATA2, GATA4, GLIS1, HMX3, HOXA7, IRF6, LMX1A, MEF2D, MEIS2, MNX1, NR2F6, NR5A1, PAX4, PDX1, PHOX2B, POU5F1(OCT4), PPARA, PTPMT1, RARA, SHOX2, SOCS4, SOX4, SPR, STAT6, TAL2, ZNF503
ccRCC-pRCC common TFs	ccRCC-chRCC common TFs	pRCC-chRCC common TFs
E2F4, IRF8	ARNTL, ARX, MNX1, POU5F1(OCT4), ZNF503	SOX4

ccRCC, clear cell renal cell carcinoma; chRCC, chromophobe renal cell carcinoma; pRCC, papillary renal cell carcinoma; TFs, transcription factors.

Since their target genes were enriched in two prominent ccRCC-related pathways—focal adhesion and steroid hormone synthesis—MYB and MYBL2 TFs were highlighted as significant TFs. Another important TF that differentially regulates the CYP7A1 gene in the context of ccRCC has been identified as the androgen receptor (AR). PDX1, MNX1, and PAX4 are TFs that enriched in maturity-onset diabetes of the young pathway and conspicuous TFs whose interaction with target genes was changed in the disease state for chRCC (Fig. 3E). GATA2, GATA4, and SHOX2 were shown to be the common TFs targeting all three genes (HTR4, GNGT1, and KCNN2) in the serotonergic synaptic pathway, which was found as disrupted in chRCC (Fig. 3F).

The TFs E2F4 and interferon regulatory factor 8 (IRF8) were discovered to be shared essential TFs for the ccRCC and pRCC subtypes. The DIFFREG method identified ARNTL, ARX, MNX1, POU5F1 (OCT4), and ZNF503 as common TFs for ccRCC and chRCC. Using the DIFFREG method, it was discovered that SOX4 is a mutually important TF for both pRCC and chRCC.

Key pathways

The KEGG pathways of target genes were analyzed using DAVID to find out functional annotations associated with target genes (Fig. 4). The DAVID online software analyzed the potential pathways of target genes with p-values <0.05. The target genes were mainly involved in focal adhesion (p = 3.2 × 10⁻²) and steroid hormone synthesis (p = 3.6 × 10⁻²) in the ccRCC. Functional enrichment analysis of target proteins in the chRCC regulatory network indicated several biological pathways mainly associated with insulin secretion (p = 1.8 × 10⁻³), maturity-onset diabetes of the young (p = 7.2 × 10⁻³), circadian entrainment (p = 3.1 × 10⁻²), and serotonergic synapse (p = 4.1 × 10⁻²). Proximal tubule bicarbonate reclamation (p = 2.1 × 10⁻⁵) and aldosterone-regulated sodium reabsorption (p = 3.2 × 10⁻²) were the most affected pathways in regard to regulatory network in pRCC cases.

FIG. 4.

The associated KEGG pathways of the target genes. The target genes were mainly involved in focal adhesion and steroid hormone synthesis in the ccRCC profile. Functional enrichment analysis of target proteins in the chRCC regulatory network indicated several biological pathways mainly associated with insulin secretion, maturity-onset diabetes of the young, circadian entrainment, and serotonergic synapse pathways. Proximal tubule bicarbonate reclamation and aldosterone-regulated sodium reabsorption pathways were the most affected pathways in regard to regulatory network in pRCC cases. KEGG, Kyoto Encyclopedia of Gene and Genome.

Discussion

Cancer biology is a culmination of multiple and interacting pathways and molecules that are under complex regulation. New approaches that consider not only particular molecular targets and pathways but also their differential regulation would offer new strides toward a deeper understanding of cancer biology, diagnosis, prognosis, and clinical management. A regulatory lens on human biology in health and disease can be achieved by harnessing the high-throughput omics technologies and systems biology (Aydin et al., 2021; Emmert-Streib et al., 2014; Hsieh et al., 2018; Kori et al., 2021).

However, most studies on biomarkers and drug targets focus on the expression level of genes, independent of mutations or post-translational modifications, which can alter the function of TFs without changing their expression level. This may result in overlooking TFs that play critical roles in disease. Because of these limitations, we applied, in this study, the DIFFREG approach to identify and prioritize the key transcriptional regulators that have a strong impact on the expression of a specific group of genes for each RCC subtype.

Key TFs and their target genes were demonstrated for each RCC subtype, also considering their subtype specificity in Figure 3. The target genes were used in the enrichment analysis to identify the disrupted signaling pathways for each subtype. The tumor specificity of the enriched pathways varied by subtype (Fig. 4). Functional enrichment analysis of proteins in the ccRCC regulome network indicated several biological pathways mainly related to focal adhesion and steroid hormone synthesis (Fig. 4). Five genes, including TNC, VASP, PDGFRB, ACTN4, and PPP1CB, were found to be implicated in the focal adhesion pathway.

Regulation of these five target genes by MYB and MYBL2 TFs was found to be differential in ccRCC patients compared with healthy samples (Fig. 3B). The MYB gene family was significantly associated with poor prognosis, proliferation, invasion, and aggressiveness of various tumors, including ccRCC (Cicirò and Sala, 2021; Sala and Ciciro, 2021). In light of the above, the molecular mechanism underlying the promoting role of the MYB gene family in tumor aggressiveness, metastasis, and proliferation could be explained by the differential regulation of genes (TNC, VASP, PDGFRB, ACTN4, and PPP1CB) involved in the focal adhesion pathway by members of the MYB gene family in ccRCC patients.

Steroid hormone synthesis was found to be another important enriched metabolic pathway in ccRCC. Steroid hormones and their receptors play an important role in normal renal biology. Three genes (UGT2B11, CYP7A1, and HSD17B8) enriched in the steroid hormone synthesis pathway were found to be dysregulated in the differential regulatory network of ccRCC. MYB and MYBL2 were determined to be key TFs that also function as differential regulators of these genes in this pathway. The AR was found to be another prominent key TF that differentially regulates the CYP7A1 gene in the case of ccRCC. Our previous study, in which we have applied a reporter features algorithm (Caliskan et al., 2020), also has revealed AR as a significant TF in ccRCC cases.

Some studies associate AR with RCC types, but they indicate that the function of AR in RCC needs to be further investigated to establish a proper correlation between AR and RCC progression (Bennett et al., 2014). The CYP7A1 protein catalyzes the first reaction of cholesterol degradation in the liver, in which cholesterol is converted to bile acids (Chiang and Ferrell, 2020). We found that AR and CYP7A1 have an activated interaction in a disease situation, leading to the downregulation of CYP7A1. This situation may lead to the accumulation of cholesterol. ccRCC is characterized by the accumulation of cholesterol, cholesterol esters, and other neutral lipids (Drabkin and Gemmill, 2012). All in all, we may conclude that the accumulation of cholesterol during the progression of ccRCC occurs through the downregulation of the CYP7A1 gene by AR (Fig. 3C).

Analysis of the functional enrichment of proteins in the chRCC regulome network indicated several biological pathways mainly associated with insulin secretion, maturity-onset diabetes of the young, circadian entrainment, and serotonergic synapse (Fig. 4). In our previous study, the insulin secretion pathway was enriched for chRCC (Caliskan et al., 2020), as well and our new findings support previous results. It has been reported that diabetes and RCCs have been listed among the risk factors for hyperglycemia (Caliskan et al., 2020). Another notable study indicated that type 2 diabetes was found to be associated with a significantly higher risk of RCC in women, independent of obesity, hypertension, and smoking. The association was strongest for non-ccRCC (Graff et al., 2018). However, they did not specify the type of non-ccRCC, and, based on our current and previous studies, we claim that chRCC differentiates from other subtypes due to its strong association with diabetes.

PDX1, MNX1, and PAX4 are TFs enriched in maturity-onset diabetes of the young pathway and conspicuous TFs whose interaction with target genes were changed in the disease state (Fig. 3E). These results suggest that genes related to diabetes are involved in the activity of the regulome network in chRCC patients. Another significantly enriched pathway in chRCC is the circadian entrainment pathway. Genes involved in the circadian clock system play an important role in biological processes such as DNA damage and repair, cell proliferation, and metastasis, thus influencing tumorigenesis and progression (Zhou et al., 2020). In addition, there is growing evidence of the importance of circadian clock genes in the diagnosis, treatment, and prognosis of cancer. Another prominent pathway is a serotonergic synapse in chRCC patients. Several studies have shown the growth-promoting effect of serotonin on various types of cancer and carcinoids (Balakrishna et al., 2021).

The common TFs targeting all three genes (HTR4, GNGT1, and KCNN2) in the serotonergic synaptic pathway were determined to be GATA2, GATA4, and SHOX2 (Fig. 3F). The TF GATA2 has been associated with poor prognosis of RCC subtypes (Peters et al., 2015), but to our knowledge, the mechanism of action has not been elucidated. Based on our findings, it can be suggested that GATA2 promotes tumor progression through alterations in the regulation of the serotonergic synaptic pathway. Moreover, in the findings of reporter features algorithms, GATA2 was also identified as a major TF in chRCC. The association with GATA4 and SHOX2 TFs and RCC has not been reported in the literature, and to our knowledge, it was reported in this study for the first time.

Pathway enrichment analysis indicated that the proximal tubule bicarbonate reclamation and aldosterone-regulated sodium reabsorption were the most influenced pathways in terms of regulatory networks in pRCC cases (Fig. 4). Among the altered pathways of the pRCC regulome network, one of the prominent pathways is proximal tubule bicarbonate reclamation, which is strongly associated with proximal tubule structure and has been previously associated with pRCC (Huang et al., 2017). Another striking result of the pRCC regulome network enrichment is the disruption of the aldosterone-regulated sodium reabsorption pathway.

Aldosterone stimulates sodium transporters in the aldosterone-sensitive distal nephron, resulting in increased sodium reabsorption, a process that also permits potassium secretion (Alexander and Bockenhauer, 2016) in healthy functioning kidneys (Solocinski and Gumz, 2015). Specific pathologies, which prevent bicarbonate reclamation and lead to metabolic acidosis, are referred to as renal tubular acidosis. It is known that these forms of renal tubular acidosis are typically accompanied by hypokalemia. Consequently, it can be concluded that these two pathways are highly connected and deterioration of the aldosterone-regulated sodium reabsorption pathway may lead to renal tubular acidosis by resulting in disruption of the bicarbonate reclamation pathway in pRCC patients.

E2F4 and IRF8 TFs were found as mutual key TFs for both ccRCC and pRCC. Our previous study, which we have applied a reporter features algorithm, also has found E2F4 as a significant TF in ccRCC. E2F4 is known as a negative regulator of VHL in the human placenta (Alahari et al., 2018). It is known that loss of the VHL gene has been strongly associated with the formation and proliferation of ccRCC through its downstream targets. A study focused on the prognostic capability of the E2F family has shown that E2F4 can be used as a prognostic marker for survival of ccRCC patients (Zhang et al., 2021), but E2F4 has not been associated with the pRCC subtype specifically.

Considering the proliferative and prognostic role of E2F4 in various tumor types and our findings, we argue that it may play an important role in the formation and proliferation of pRCC by negatively regulating its downstream targets. IRF8 has been proposed as a predictor of progression and patient survival in RCC (Muhitch et al., 2019), but it has not been associated with the pRCC subtype specifically. Considering our findings, it can be concluded that IRF8 has been implicated in both ccRCC and pRCC formation and progression.

ARNTL, ARX, MNX1, POU5F1, and ZNF503 were determined as a mutual key TF for both ccRCC and chRCC through the DIFFREG algorithm. The stem cell factor POU5F1 has been correlated with advanced tumor stage and poorer overall survival in RCC patients (Siebenthall et al., 2019). The ARX gene is a homeobox-containing gene expressed during development and functions in the nervous system and brain development. However, intriguing studies of the role of ARX orthologs in diverse models have suggested that it is critical for important developmental processes such as proliferation and cell differentiation (Friocourt et al., 2006) but has not been correlated with RCC. This study is the first time that ARX and ZNF503 are proposed as potential biomarkers for both ccRCC and chRCC.

SOX4 was found as a mutual significant TF for both pRCC and chRCC through the DIFFREG approach. The study focusing on the role of SOX4 in RCC indicated that specific knockdown of SOX4 in renal cancer cell lines significantly suppressed the migration and invasion of cancer cells; specific overexpression of SOX4 in renal epithelial cell line markedly promoted the migration and invasion of the cell line (Ruan et al., 2017).

Our previous study, in which we have applied reporter features algorithm, (Caliskan et al., 2020) also has revealed significant TFs specific to RCC subtypes. AR, E2F4, ESR1, and NR1I2 are common highlighted TFs that both DIFFREG and reporter feature algorithm gave a result for ccRCC cases. ESR1 is an estrogen receptor that is known as an oncogene and has been strongly associated with RCC.

There is significant evidence that these receptors promote the development and progression of a variety of cancers, with new research suggesting that the human kidney could be one of them. This is supported by clinical evidence, such as a large sex difference in RCC incidence, with men having double the rate of women. Furthermore, the study focusing on hormone therapy of RCC in hamsters led to the concept that some kidney cancers are hormone dependent (El-Deek et al., 2018). E2F4 and YBX1 became prominent in both DIFFREG and reporter feature algorithms in pRCC samples. The relationship of E2F4, YBX1, and pRCC has been discussed in detail in the previous paragraphs. Only GATA2 was determined to be mutual for chRCC in both the DIFFREG and reporter feature algorithms.

Conclusion

Understanding the dynamics of cellular systems and disease molecular pathways requires the identification of differential regulators. Most biomarker or drug delivery molecule research focuses on gene expression levels, ignoring mutations or post-translational modifications, which can change TF function without changing their expression levels. Considering these arguments, we developed the DIFFREG algorithm, which allows us to identify important TFs that may not be DEG. The DIFFREG algorithm was applied to three subtypes of RCC in this study; however, the algorithm is applicable to transcriptome data of any illness or to other organisms for which TF-target gene data are available. These analyses address the question of which TF-gene interactions are perturbed between the disease and normal conditions and which regulators tend to alter their targets.

The differential regulome methodology focuses on the differential regulation between TFs and their target genes rather than differential expression. We focused on functional enrichment analyses and highlighted signaling pathways when evaluating the results of this methodology. Focal adhesion and steroid hormone synthesis were highlighted in ccRCC cases, while proximal tubule bicarbonate reclamation and aldosterone-regulated sodium reabsorption were emphasized in pRCC patients. Insulin secretion, maturity-onset diabetes of the young, circadian entrainment, and serotonergic synapse were found as most influenced ones in chRCC disease.

The discovery of these subtype-specific biomarkers may play an important role in elucidating the molecular mechanisms of RCC development and translating the knowledge of the genetic basis of RCC into the clinic. Furthermore, these indicators may aid oncologists in making the best diagnosis and prognosis management decisions.

Footnotes

Authors' Contributions

A.C.: Writing—original draft; conceptualization; software; methodology; and visualization. K.Y.A.: Conceptualization; writing—review, and editing; methodology; and supervision. All authors read and approved the final version of the article.

Author Disclosure Statement

The authors declare they have no conflicting financial interests.

Funding Information

No specific funding was received for this work.

Supplementary Material

Abbreviations Used

References

Alahari

, Garcia

, Post

, et al. The von Hippel Lindau tumour suppressor gene is a novel target of E2F4-mediated transcriptional repression in preeclampsia. Biochim Biophys Acta Mol Basis Dis, 2018; 1864(10):3298–3308; doi: 10.1016/j.bbadis.2018.07.016

Alexander

, Bockenhauer

. Renal tubular acidosis. Pediatr kidney Dis, 2016; 973–991; doi: 10.1016/j.pcl.2018.08.011

Anders

, Huber

. Differential expression analysis for sequence count data. Genome Biol, 2010; 11(10):R106; doi: 10.1186/gb-2010-11-10-r106

Aydin

, Caliskan

, Arga

. Overview of omics biomarkers in pituitary neuroendocrine tumors to design future diagnosis and treatment strategies. EPMA J, 2021; 12(3):383–401; doi: 10.1007/s13167-021-00246-1

Balakrishna

, George

, Hatoum

, et al. Serotonin pathway in cancer. Int J Mol Sci, 2021; 22(3):1268; doi: 10.3390/ijms22031268

Bennett

, Rajandram

, Ng

, et al. Evaluation of steroid hormones and their receptors in development and progression of renal cell carcinoma. J Kidney Cancer VHL, 2014; 1(2):17; doi: 10.15586/jkcvhl.2014.9

Bovolenta

, Acencio

, Lemke

. HTRIdb: An open-access database for experimentally verified human transcriptional regulation interactions. BMC Genom, 2012; 13(1):405; doi: 10.1186/1471-2164-13-405

Caliskan

, Andac

, Arga

. Novel molecular signatures and potential therapeutics in renal cell carcinomas: Insights from a comparative analysis of subtypes. Genomics, 2020; 112(5):3166–3178; doi: 10.1016/j.ygeno.2020.06.003

Caliskan

, Gulfidan

, Sinha

, et al. Differential interactome proposes subtype-specific biomarkers and potential therapeutics in renal cell carcinomas. J Pers Med, 2021; 11(2):158; doi: 10.3390/jpm11020158

10.

Cao

, Luo

, Cheng

, et al. Three-dimensional regulation of transcription. Protein Cell, 2015; 6(4):241–253; doi: 10.1007/s13238-015-0135-7

11.

Chiang

JYL

, Ferrell

. Up to date on cholesterol 7 alpha-hydroxylase (CYP7A1) in bile acid synthesis. Liver Res, 2020; 4(2):47–63; doi: 10.1016/j.livres.2020.05.001

12.

Cicirò

, Sala

. MYB oncoproteins: Emerging players and potential therapeutic targets in human cancer. Oncogenesis, 2021; 10(2):1–15; doi: 10.1038/s41389-021-00309-y

13.

Dennis

, Sherman

, Hosack

, et al. DAVID: Database for annotation, visualization, and integrated discovery. Genome Biol, 2003; 4(9):1–11; doi: 10.1186/gb-2003-4-9-r60

14.

Drabkin

, Gemmill

. Cholesterol and the development of clear-cell renal carcinoma. Curr Opin Pharmacol, 2012; 12(6):742–750; doi: 10.1016/j.coph.2012.08.002

15.

El-Deek

HEM

, Ahmed

, Hassan

, et al. Expression and localization of estrogen receptors in human renal cell carcinoma and their clinical significance. Int J Clin Exp Pathol, 2018; 11(6):3176.

16.

Emmert-Streib

, Dehmer

, Haibe-Kains

. Gene regulatory networks and their applications: Understanding biological and medical problems in terms of networks. Front Cell Dev Biol, 2014; 2:38; doi: 10.3389/fcell.2014.00038

17.

Essaghir

, Pachikian

, Delzenne

, et al. TFactS: A Tool to Predict Transcription Factor Regulation from Gene Expression Data. In: The 9th Conference on Computational Biology (ECCB10), Ghent-Belgium. 2010.

18.

Friocourt

, Poirier

, Rakić

, et al. The role of ARX in cortical development. Eur J Neurosci, 2006; 23(4):869–876; doi: 10.1111/j.1460-9568.2006.04629.x

19.

Graff

, Sanchez

, Tobias

, et al. Type 2 diabetes in relation to the risk of renal cell carcinoma among men and women in two large prospective cohort studies. Diabetes Care, 2018; 41(7):1432–1437; doi: 10.2337/dc17-2518

20.

Han

, Shim

, Shin

, et al. TRRUST: A reference database of human transcriptional regulatory interactions. Sci Rep, 2015; 5:11432; doi: 10.1038/srep11432

21.

Hsieh

, Le

, Cao

, et al. Genomic classifications of renal cell carcinoma: A critical step towards the future application of personalized kidney cancer care with pan-omics precision. J Pathol, 2018; 244(5):525–537; doi: 10.1002/path.5022

22.

Huang

, Yuan

, Wu

, et al. An integrated analysis for long noncoding RNAs and microRNAs with the mediated competing endogenous RNA network in papillary renal cell carcinoma. Onco Targets Ther, 2017; 10:4037; doi: 10.2147/OTT.S141951

23.

Kori

, Aydin

, Gulfidan

, et al. The repertoire of glycan alterations and glycoproteins in human cancers. Omi A J Integr Biol, 2021; 25(3):139–168; doi: 10.1089/omi.2020.0210

24.

Liu

Z-P

, Wu

, Miao

, et al. RegNetwork: An integrated database of transcriptional and post-transcriptional regulatory networks in human and mouse. Database, 2015; 2015:1–12; doi: 10.1093/database/bav095

25.

Muhitch

, Hoffend

, Azabdaftari

, et al. Tumor-associated macrophage expression of interferon regulatory factor-8 (IRF8) is a predictor of progression and patient survival in renal cell carcinoma. J Immunother cancer, 2019; 7(1):1–9; doi: 10.1186/s40425-019-0630-0

26.

Peters

, Dubrowinskaja

, Tezval

, et al. Decreased mRNA expression of GATA1 and GATA2 is associated with tumor aggressiveness and poor outcome in clear cell renal cell carcinoma. Target Oncol, 2015; 10(2):267–275; doi: 10.1007/s11523-014-0335-8

27.

Portales-Casamar

, Arenillas

, Lim

, et al. The PAZAR database of gene regulatory information coupled to the ORCA toolkit for the study of regulatory sequences. Nucleic Acids Res, 2008; 37(suppl_1):D54–D60; doi: 10.1093/nar/gkn783

28.

Ruan

, Yang

, Wei

, et al. Overexpression of SOX4 promotes cell migration and invasion of renal cell carcinoma by inducing epithelial-mesenchymal transition. Int J Oncol, 2017; 51(1):336–346; doi: 10.3892/ijo.2017.4010

29.

Sala

, Ciciro

. MYB oncoproteins: Emerging players and potential therapeutic targets in human cancer. Oncogenesis, 2021; 10(2):1–15; doi: 10.1038/s41389-021-00309-y

30.

Siebenthall

, Miller

, Vierstra

, et al. Integrated epigenomic profiling reveals endogenous retrovirus reactivation in renal cell carcinoma. EBioMedicine, 2019; 41:427–442; doi: 10.1016/j.ebiom.2019.01.063

31.

Solocinski

, Gumz

. The circadian clock in the regulation of renal rhythms. J Biol Rhythms, 2015; 30(6):470–486; doi: 10.1177/0748730415610879

32.

Del Vecchio

, Mastroiaco

, Di Marco

, et al. Next-generation sequencing: Recent applications to the analysis of colorectal cancer. J Transl Med, 2017; 15(1):246; doi: 10.1186/s12967-017-1353-y

33.

Verfaillie

, Imrichová

, Van de Sande

, et al. IRegulon: From a gene list to a gene regulatory network using large motif and track collections. PLoS Comput Biol, 2014; 10(7):e1003731; doi: 10.1371/journal.pcbi.1003731

34.

Zhang

, Cui

, Wang

, et al. Comprehensive analysis of the expression and prognosis for E2Fs in human clear cell renal cell carcinoma. J Healthc Eng, 2021; 2021:5790416; doi: 10.1155/2021/5790416

35.

Zhou

, Luo

, Li

, et al. Circadian clock is associated with tumor microenvironment in kidney renal clear cell carcinoma. Aging (Albany, NY), 2020; 12(14):14620; doi: 10.18632/aging.103509

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.16 MB