Archives of Gynecology and Obstetrics

, Volume 290, Issue 4, pp 749–755

Construction of breast cancer gene regulatory networks and drug target optimization

Gynecologic Oncology

DOI: 10.1007/s00404-014-3264-y

Cite this article as:
Xie, Y., Wang, R. & Zhu, J. Arch Gynecol Obstet (2014) 290: 749. doi:10.1007/s00404-014-3264-y

Abstract

Objective

The purpose of this study was to construct the breast cancer gene regulatory networks through the high-throughput techniques and optimize the drug target genes of breast cancer using bioinformatics analysis, and thus accelerate the process of drug development and improve the cure rate of breast cancer.

Methods

The gene expression profile data of breast cancer were downloaded from GEO database and the transcriptional regulation data were obtained from UCSC database. Then we identified the differentially expressed genes (DEGs) by SAM algorithm and built gene regulatory networks by the supervised algorithm SIRENE. Finally, the drug targets of the DEGs with changed regulation relations were optimized based on the CancerResource database.

Results

A total of 584 DEGs were identified and the gene regulatory networks in the normal state and tumorous state were constructed. By comparing the new predicted regulatory relation in cancer state and normal state, the regulatory relation of 18 genes was found to be changed in the two states, showing the possibility to be applied as drug target genes. After the searches in the CancerResources, 7 genes were screened as the drug target genes, such as PFKFB3.

Conclusion

Our present findings shed new light on the molecular mechanism of breast cancer and provide some drug targets which have the potential to be used in clinic for the treatment of breast cancer in future.

Keywords

Breast cancer Differentially expressed gene Gene regulatory network Target gene optimization 

Introduction

Breast cancer is one of the most common gynecological malignancies and has the highest mortality of cancer in American women [1]. In recent years, the morbidity of breast cancer increased gradually, accounting for the largest percentage in gynecological malignancies [2]. The treatment effect of breast cancer has a great association with the stages in which patients are diagnosed [3, 4]. With the development of biotechnology, our understanding of the pathogenesis of breast cancer is improving and genes related to breast cancer have been constantly found as well [5]. It has become a research hotspot to find more accurate staging and classification genetic markers, diagnostic and prognostic markers from molecular level to improve the cure rate of breast cancer [6, 7]. The genetic therapy has become an important part in the treatment of tumor biology and also has shown great application value in the breast cancer treatment [8].

Using the comprehensive bioinformatics method, multiple genes have been found to be differentially expressed in breast cancer, such as HER2, TGFBR2, CXCL12, and MMP3. In addition, the combination of molecular network analysis technology and gene expression profile in bioinformatics has shown significant application prospect to classify various diseases and identify new therapeutic targets [9]. For example, by comparing the gene expression patterns of metastatic and non-metastatic breast cancer tumor tissue, special protein components were identified in the protein interaction network as the risk marker of breast cancer metastasis [10]. In this study, we aimed to investigate the changes of regulatory relations between the DEGs and the corresponding transcription factors, so as to identify the genes involved in the relations. Drug target related with the selected genes in breast cancer were also optimized.

Materials and methods

Gene expression profile obtaining and pretreatment

The Gene expression omnibus (GEO, http://www.ncbi.nlm.nih.gov/geo/) database is the largest and most comprehensive public gene expression data resource. The gene expression data stored in GEO database are submitted by laboratories around the world, including Affymetrix microarray data, EST sequence data, SAGE sequence data and second-generation sequencing data [11]. Each set of gene expression profile data includes platform information, family information and sample information. We downloaded the gene expression profile data of breast cancer (GSE15852) from GEO database [12], including 43 normal tissue samples and 43 breast cancer tissue samples. Through the corresponding relation between the probe and gene in microarray platform GPL96, we translated the probe number in the expression profile into corresponding gene symbol, and the average expression value was adopted as the representative of the genes that have multiple expression values. After pretreatment, we got 13,784 genes totally.

Transcriptional regulation data obtaining

The only way to achieve every activity of cells is the real-time expression detection of specific genes and the specific transcription factors. Generally, the disturbed regulation of gene expression will lead to cell dysfunction, even the occurrence of cancers. We downloaded the data of human h18 transcription factor binding sites and the information of gene coordinate location from UCSC database [13], to screen the regulatory relations.

Identification of the differentially expressed genes

After the standardization of Affymetrix data, we used the significance analysis of microarrays (SAM) algorithm to screen the DEGs in the breast cancer tissue samples compared with the normal tissue samples [14]. SAM algorithm is a mature method to select the exact genes, which can correct the false positive rate in multiple testing with large gene numbers by controlling the FDR (false discovery rate).
$${\text{Relative}}\, {\text{difference}} \,{\text{statistics}} \;d = \frac{{X{'_1} - X{'_2}}}{{S + {S_0}}}$$

The statistics d measured the relative difference in gene expression and was the correction of t-statistic. X1′ represents the average gene expression level in State 1. X2′ represents the average gene expression level in State 2. S represents the gene variance and S0 represents deviation correction quantity.

Construction of gene regulatory networks

Previous studies obtained the regulatory relation between transcription factor and target genes through the sequence information. With the improvement of high-throughput methods and techniques, we could use gene expression profile to speculate the gene regulatory relation with high-throughput data, and the algorithms CLR [15], RN [16] and ARACNE [17] based on mutual information, as well as the gene correlation algorithms WGCNA [18], CORRELATION [19] and the algorithms GENIE3 [20] based on random walk are the commonly used methods. SIRENE [17] based on support vector machine (SVM) is reported to have higher efficiency compared to the above method [21]. In the algorithm SIRENE, the target genes of known transcription factor are regarded as positive training set and non-target genes are regarded as negative training set, where the gene expression value is used as input and the target (0) or non-target (1) is used as output, while the sample volume is used as the characteristics. Each transcription factor trained one model and this model was used to predict whether the specific gene was the target gene of the transcription factor or not.

We inputted the expression values of DEGs and the regulatory relations screened from the UCSC database to SIRENE, to construct the gene regulatory networks with the help of Cytoscape [22]. What is more, with the expression values of DEGs in normal and cancer tissue samples, we got two regulatory networks: GRN_norm, GRN_tumor.

Drug targets identification

The target genes whose regulatory relation was different in the normal state and diseased state were more likely to be associated with the occurrence of breast cancer. They have the opportunities to cause cancers in the change of regulatory relation. Thus, to screen the genes which could be used as drug targets, genes with the changed regulatory relations based on the gene regulatory network were further mapped to the CancerResource database [23].

Results

Identification of differentially expressed genes

In this study, under the set-off criteria of fold-change = 1.5 and FDR = 0.05, we identified 584 DEGs, including 124 up-regulated genes and 460 down-regulated genes.

Construction of gene regulatory networks

A total of 214,608 pairs of regulatory relation between 216 transcription factors and 16,863 genes were obtained from the UCSC database. Among them, there were 442 pairs of transcriptional regulatory relations of the DEGs. The expression values of DEGs and regulatory relations were inputted to the SIRENE as priori knowledge.

The GRN-NORM regulatory network contained 467 pairs of transcriptional regulation information (Fig. 1), of which 26 pairs were newly predicted regulatory relations (Table 1).
Fig. 1

Gene regulatory network in normal state. The green nodes represent transcription factor (TF) and red nodes represent target gene. The green lines show the predicted regulatory relation and gray lines show the regulatory relation existed in UCSC

Table 1

New predicted regulatory relations in normal state (GRN-NORM)

TF

Target

TF

Target

CEBPA

FABP4

FOSB

TJP2

CEBPA

ITGA7

HLF

ITGA7

CEBPB

FMOD

JUN

BCL2L2

CEBPB

ITGA7

JUN

CD81

CEBPB

PFKFB3

JUN

FMOD

EPAS1

ITGA7

JUN

MAN1A1

FOS

BCL2L2

JUN

S100A10

FOS

FMOD

JUN

KLF4

FOS

MAN1A1

JUN

TJP2

FOS

TJP2

JUN

PRDX6

FOSB

BCL2L2

STAT5A

ADFP

FOSB

FMOD

STAT5A

BNIP3

FOSB

MAN1A1

STAT5A

FMOD

TF transcription factors

The GRN-TUMOR regulatory network contained 462 pairs of transcriptional regulation information (Fig. 2), of which 22 pairs were the newly predicted regulatory relation (Table 2).
Fig. 2

Gene regulatory network in tumorous state. The green nodes represent transcription factor (TF) and red nodes represent target gene. The green lines show the predicted regulatory relation and gray lines show the regulatory relation existed in UCSC

Table 2

New predicted regulatory relations in tumorous state (GRN-TUMOR)

TF

Target

TF

Target

CEBPA

ITGA7

JUN

BCL2L2

CEBPB

ITGA7

JUN

CDKN1C

EPAS1

ITGA7

JUN

DPT

FOS

ATP1A2

JUN

LEP

FOS

FOXO1

JUN

LGALS3

FOS

TJP2

JUN

P2RY1

FOSB

ATP1A2

JUN

S100A10

FOSB

FOXO1

JUN

CSDE1

FOSB

TJP2

JUN

TJP2

HLF

ITGA7

JUN

PRDX6

JUN

ATP1A2

JUN

NR1H3

TF transcription factors

Comparison and analysis of gene regulatory network

By analyzing the newly predicted regulatory relations in cancer state and normal state, we found that 10 pairs of regulatory relation existed in the both states, and the other 38 pairs of regulatory relation were specific. What is more, the regulatory relation of four genes in normal was consistent with that in tumorous state, as shown in Fig. 3.
Fig. 3

New predicted regulatory relation. The green, red and yellow nodes represent TF, target genes and the target genes whose regulatory relation was unchanged, respectively. The light blue, dark blue and red lines represent the regulatory relation existed in normal state, tumorous state and both state, respectively

Drug targets

We searched 18 target genes, whose regulatory relation was changed, in the drug target database CancerResource, and found that seven genes could be used as drug targets, as shown in Table 3. In these 7 genes, PFKFB3 is a known target gene of breast cancer, while the other 11 genes should be studied further as candidate drug targets of breast cancer.
Table 3

The information of the 7 genes in CancerResource database

 

PFKFB3

BNIP3

CDKN1C

FOXO1

MAN1A1

KLF4

ATP1A2

Salinomycin

      

Decitabine

 

   

Oxaliplatin

    

  

Bicalutamide

    

  

Vorinostat

   

   

Fluorouracil

    

 

Oxygen

 

    

Quercetin

  

    

Adaphostin

   

   

Digitoxin

      

Deoxycholic acid

      

Ouabain

      

Cholan-24-oic acid, 3,12-dihydroxy-, (3.alpha., 5.beta., 12.alpha.)-, sodium salt ···

      

3-Beta-d-ribofuranosyl-1H-pyrrole-2,5-dione

      

Edelfosine

      

Ilmofosine

      

Seocalcitol

  

    

Discussion

In this study, a total of 584 DEGs were identified, including 124 up-regulated genes and 460 down-regulated genes. The expression values of these DEGs were applied to build gene regulatory networks in normal state and tumorous state, respectively. In the normal gene regulatory network, 467 pairs of transcriptional regulatory relation were displayed, including 26 pairs of newly predicted relations; and in the diseased gene regulatory network, 462 pairs of transcriptional regulatory relation were displayed, including 22 pairs of newly predicted relations. In the 48 pairs of new predicted regulatory relation, 10 pairs existed in the two states at the same time and the other 38 pairs were specific.

Normal cell activities are among the most important factors to keep human body healthy, it is able to detect every activity of cells by detecting the gene expression values. What is more, gene expression is controlled by specific transcription factor. So the changes in the regulatory relation between genes and transcription factor may lead to dysfunction of cells and finally cause cancers. Therefore, the DEGs whose regulatory relation was different in the normal tissue and breast cancer tissue are possible to be used as drug target genes.

By comparing the newly predicted regulatory relation in cancer state and normal state, we found that the 18 genes with changed regulatory relation could potentially be applied as drug target genes. Among these 18 genes, there were seven known drug target genes which have been approved by FDA. As shown in Table 3, lots of small-molecular drugs, such as salinomycin and fluorouracil, and some small molecules, such as oxygen, can influence the gene regulatory relation of the seven target genes and thus kill the cancer cells in human body. For example, salinomycin, a polyether antibiotic, is the leading drug for the treatment of breast cancer and has been recognized as novel and effective anti-cancer agent [24]. The main effect of the salinomycin on the target gene PFKFB3 is to induce massive apoptosis in human cancer cells but not in normal cells [25, 26]. Therefore, it is reasonable to believe that the other 11 genes could be used as new drug target for the treatment of breast cancer hopefully. Further biochemical experiments displaying their functions are needed to analyze and confirm more activities and features of the 11 genes, and then develop novel drugs for the treatment of breast cancer. Besides the small-molecular drugs, variety of novel anti-cancer agents are now been increasingly recognized as target drugs for breast cancer, such as monoclonal antibody drugs and aromatase inhibitors [27, 28]. They may also have positive effect on the new predicted target genes to control breast cancer.

The role of gene regulatory networks has been extensively mentioned in relation to breast cancer. It has been claimed that the effect of the gene interaction networks may be much more important than individual gene contributions [29]. However, the regulation networks are likely comprised of complex genetic interactions and thus it makes more difficulties to understand the mechanisms of gene regulation during breast cancer [30]. In the gene regulatory network, the same transcription factor may control the expression of hundred genes and a single gene may be controlled by several regulatory factors [31]. Some studies have proved that several transcription factors that supplemented with gene regulatory networks act as main master regulators in primary breast cancer development [32]. It provided a potential pharmacological method for the treatment of breast cancer through the application of drugs that targeted specific genes.

With the development of biotechnology and the improvement of methodology, the priori knowledge will be accumulated continuously and various omics data will increase as well. More accurate methods for the prediction of gene regulatory network will be found eventually. We can believe that more and more novel drug target genes of breast cancer will be identified.

Conflict of interest

None.

Copyright information

© Springer-Verlag Berlin Heidelberg 2014

Authors and Affiliations

  1. 1.Department of OncologyRenmin Hospital of Wuhan UniversityWuchanChina
  2. 2.Department of EmergencyRenmin Hospital of Wuhan UniversityWuchanChina
  3. 3.General Surgery DepartmentXin Hua Hospital Affiliated to Shanghai Jiao Tong University School of MedicineShanghaiChina

Personalised recommendations