Integrated analysis of differentially expressed genes and pathways in triple-negative breast cancer

Abstract

Triple-negative breast cancer (TNBC) is a heterogeneous disease characterized by an aggressive phenotype and reduced survival. The aim of the present study was to investigate the molecular mechanisms involved in the carcinogenesis of TNBC and to identify novel target molecules for therapy. The differentially expressed genes (DEGs) in TNBC and normal adjacent tissue were assessed by analyzing the GSE41970 microarray data using Qlucore Omics Explorer, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes. Pathway enrichment analyses for DEGs were performed using the Database for Annotation, Visualization and Integrated Discovery online resource. A protein-protein interaction (PPI) network was constructed using Search Tool for the Retrieval of Interacting Genes, and subnetworks were analyzed by ClusterONE. The PPI network and subnetworks were visualized using Cytoscape software. A total of 121 DEGs were obtained, of which 101 were upregulated and 20 were downregulated. The upregulated DEGs were significantly enriched in 14 pathways and 83 GO biological processes, while the downregulated DEGs were significantly enriched in 18 GO biological processes. The PPI network with 118 nodes and 1,264 edges was constructed and three subnetworks were extracted from the entire network. The significant hub DEGs with high degrees were identified, including TP53, glyceraldehyde-3-phosphate dehydrogenase, cyclin D1, HRAS and proliferating cell nuclear antigen, which were predominantly enriched in the cell cycle pathway and pathways in cancer. A number of critical genes and pathways were revealed to be associated with TNBC. The present study may provide an improved understanding of the pathogenesis of TNBC and contribute to the development of therapeutic targets for TNBC.

Introduction

Triple-negative breast cancer (TNBC) is an aggressive subtype of breast cancer with a poor prognosis and high mortality, which is diagnosed more frequently in young and premenopausal women (1,2). Due to the absence of estrogen receptors, progesterone receptor and the human epidermal growth factor receptor-2 in TNBC tumor types, patients with TNBC do not respond to hormone- or trastuzumab-based therapies, leaving cytotoxic chemotherapy as the current therapy (3). Therefore, the identification of critical genes involved in TNBC carcinogenesis may provide a strategy for molecular target therapy of TNBC (4).

Microarray technology, which may be used to detect the global gene expression, has provided an alternative method for the molecular classification of different types of cancer, as well as the exploration of potential prognostic biomarkers and therapeutic targets. Based on microarray analysis, breast cancer has been divided into distinctive subtypes according to different gene expression patterns (5–11). Furthermore, Rakha et al (4) performed a microarray analysis on a relatively large set of 1,944 cases of invasive breast cancer that contained TNBC tissues, as well as information on tumor size, lymph node involvement and androgen receptor expression levels, and determined these three parameters as the most useful prognosticators of TNBC. Therefore, a systematic analysis of gene expression patterns in TNBC may aid researchers to comprehend the development and the treatment of this disease and to identify novel therapeutic targets.

In the present study, biological microarray analysis was used to analyze the gene expression profile of TNBC and to screen the differentially expressed genes (DEGs) between TNBC and adjacent normal tissues. Furthermore, altered biological pathways in TNBC were identified using bioinformatics tools. Additionally, a protein-protein interaction (PPI) network of DEGs was constructed in order to identify the crucial genes involved in the process of TNBC. The present study aimed to improve the understanding of the underlying pathological mechanism and facilitate the discovery of potential novel therapeutic targets for TNBC.

Materials and methods

Gene expression data

The GSE41970 gene expression profiles of TNBC were downloaded from Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo) of the National Center for Biotechnology Information (Bethesda, MD, USA) based on the GPL16299 platform data (NanoString nCounter mRNA Human Cancer Reference Kit; NanoString, Inc., Ticino, Bellinzona, Switzerland) (12). In total, 200 specimens, including 160 primary TNBC specimens and 40 normal samples, were used in the present study.

Data processing and differential analysis

Qlucore Omics Explorer (QOE) software (version 3; Qlucore AB, Lund, Sweden) was used to analyze the data from the microarray (13,14). Intensity values of each probe-set were log₂ transformed, and a gene-specific t-test was subsequently performed between TNBC samples and matched normal samples. P<0.01 and [log₂(fold change)>2] were regarded as the cut-off criteria to screen out DEGs. To generate an overview of the gene expression profile, hierarchical clustering was also performed by QOE.

Gene ontology (GO) and pathway enrichment analysis

To determine the biological pathways altered in TNBC, GO biological process terms and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses were performed on the DEGs using Database for Annotation, Visualization and Integration Discovery (version 6.7; DAVID; http://david.abcc.ncifcrf.gov) (15). P<0.01 and false discovery rate (FDR) <0.01 were set as the cut-off value.

Construction and analysis of PPI network

It is possible to utilize PPI research to study DEG protein functions at the molecular level, and cellular regulatory mechanisms can be interpreted by elucidation of genome-wide protein interactions (16). The PPI network was constructed for the DEGs using information provided by the Search Tool for the Retrieval of Interacting Genes, version 10 (STRING; http://string-db.org) (17), and was subsequently visualized using Cytoscape version 3.2.1 (http://cytoscape.org) (18). The interactions of protein pairs with a combined score >0.5 were retained in the network and the hub genes were screened according to the degree score (the number of neighbors). The subnetworks were then analyzed by Clustering with Overlapping Neighborhood Expansion (ClusterONE; http://www.paccanarolab.org/clusterone). GO functional and KEGG pathway enrichment analyses of the most significant subnetworks were performed with a threshold of P<0.01 and FDR<0.01.

Results

Identification of DEGs in TNBC

Using QOE, 121 DEGs were identified between 160 TNBC and 40 normal tissues, including 101 upregulated genes and 20 downregulated genes. The 10 most significantly up- or downregulated genes are listed in Table I.

Table I.

Top 10 most significantly up- or downregulated differentially expressed genes in triple-negative breast cancer compared with normal tissue.

Gene symbol	P-value	Fold-change
Upregulated genes
BIRC5	6.50×10−36	48.63
MYBL2	4.32×10−34	57.33
TOP2A	3.49×10−31	41.91
CDC2	4.07×10−29	32.75
MMP9	2.09×10−23	21.35
CHEK1	8.68×10−23	13.37
SPP1	2.21×10−22	40.41
TYMS	4.26×10−22	18.21
E2F1	4.38×10−21	11.18
PCNA	1.27×10−20	6.87
Downregulated genes
IGFBP6	2.19×10−14	0.12
ESR1	1.09×10−13	0.09
DLC1	1.27×10−12	0.14
EGR1	6.87×10−11	0.20
IGF1	1.72×10−10	0.27
TGFBR3	3.37×10−10	0.16
PPARG	3.84×10−10	0.23
NGFR	3.93×10−10	0.12
CD34	6.67×10−9	0.27
FOS	5.83×10−8	0.30

Clustering of DEGs

The 121 DEGs in TNBC compared with normal tissues were selected for hierarchical clustering analysis. As presented in Fig. 1, the DEGs were divided into two major groups, separating 101 upregulated genes from 20 downregulated genes.

Figure 1.

Cluster analysis of differentially expressed genes. Red highlights genes with high expression, green shows cells with low expression and black indicates no significant change in the expression level between the mean and sample.

GO of biological process enrichment analysis of up- and downregulated DEGs

The upregulated DEGs were significantly enriched in 83 biological processes and the top five significantly enriched processes were mainly associated with the regulation of cell proliferation, cell cycle and cell apoptosis (Table II). The downregulated DEGs were significantly enriched in 18 biological processes, among which the most significant were mainly involved in the regulation of cell proliferation, macromolecule metabolic process and cell migration (Table II).

Table II.

Top five significantly enriched biology processes for differentially expressed genes.

GO ID	Term	Gene counts	P-value	FDR
Upregulated
GO:0042127	Regulation of cell proliferation	35	2.22×10−18	3.76×10−15
GO:0051726	Regulation of cell cycle	24	8.20×10−17	1.89×10−13
GO:0008284	Positive regulation of cell proliferation	26	8.35×10−17	1.89×10−13
GO:0042981	Regulation of apoptosis	31	2.15×10−14	3.63×10−11
GO:0043067	Regulation of programmed cell death	31	2.79×10−14	4.72×10−11
Downregulated
GO:0042127	Regulation of cell proliferation	12	1.19×10−9	1.88×10−6
GO:0010604	Positive regulation of macromolecule metabolic process	11	5.40×10−8	8.57×10−5
GO:0030334	Regulation of cell migration	7	8.25×10−8	1.31×10−4
GO:0040012	Regulation of locomotion	7	1.76×10−7	2.79×10−4
GO:0051270	Regulation of cell motion	7	1.81×10−7	2.88×10−4

GO, gene ontology; Gene counts, number of genes; FDR, false discovery rate.

KEGG pathway enrichment analysis of up- and downregulated DEGs

The upregulated DEGs were significantly enriched in 14 pathways, which mainly involved the cell cycle and pathways associated with cancer, including the p53 signaling pathway (Table III). In the downregulated DEGs, four pathways were identified, however they were without statistical significance (P>0.01 or FDR>0.01).

Table III.

Significantly enriched pathways for differentially expressed genes.

KEGG ID	Term	Gene counts	P-value	FDR
hsa05200	Pathways in cancer	37	3.77×10−24	4.19×10−21
hsa05219	Bladder cancer	17	8.90×10−20	9.87×10−17
hsa05223	Non-small cell lung cancer	14	3.04×10−13	3.38×10−10
hsa04110	Cell cycle	18	1.03×10−12	1.14×10−9
hsa05215	Prostate cancer	16	1.05×10−12	1.16×10−9
hsa05212	Pancreatic cancer	14	1.59×10−11	1.76×10−8
hsa05220	Chronic myeloid leukemia	14	2.74×10−11	3.04×10−8
hsa05214	Glioma	13	5.32×10−11	5.91×10−8
hsa05218	Melanoma	13	2.33×10−10	2.59×10−7
hsa05213	Endometrial cancer	11	2.22×10−9	2.46×10−6
hsa05222	Small cell lung cancer	12	2.30×10−8	2.56×10−5
hsa05216	Thyroid cancer	8	1.25×10−7	1.39×10−4
hsa05210	Colorectal cancer	10	2.74×10−6	3.04×10−3
hsa04115	p53 signaling pathway	9	5.03×10−6	5.58×10−3

KEGG, Kyoto Encyclopedia of Genes and Genomes; Gene counts, number of genes; FDR, false discovery rate.

PPI network analysis

Based on STRING database analysis, a total of 1,264 protein pairs with the combined score >0.5 were identified. The PPI network including 118 nodes and 1,264 edges was constructed (Fig. 2) and the connectivity degree of each node was calculated. The top five nodes TP53 (degrees, 86), glyceraldehyde-3-phosphate dehydrogenase (GAPDH; degrees, 62), cyclin D1 (CCND1; degrees, 58), HRAS (degrees, 58) and proliferating cell nuclear antigen (PCNA; degrees, 52) with degrees >50 were screened as hub proteins in the PPI network.

Figure 2.

Protein-protein interaction network of differentially expressed genes. The red nodes represent upregulated genes and the green nodes represent downregulated genes. The lines represent the interaction between proteins.

Three subnetworks (1,2 and 3) with P<0.05 were established using the ClusterONE plugin (Fig. 3). The hub proteins TP53, GAPDH, CCND1, HRAS and PCNA were demonstrated to be involved in all of these three subnetworks. Subnetwork 1 was mainly associated with regulation of cell proliferation and cell communication, while the significant pathways were mainly associated with pathways in cancer (Tables IV and V). By contrast, the other subnetworks were mainly associated with the cell cycle (Table IV). Additionally, the significant pathways correlated with subnetwork 2 and 3 were the cell cycle, the p53 signaling pathway and pathways in cancer (Table V).

Figure 3.

Three subnetworks in the protein-protein interaction network of differentially expressed genes. (A) Subnetwork 1; (B) Subnetwork 2; and (C) Subnetwork 3 were determined. The red nodes represent upregulated genes and the green nodes represent downregulated genes. The lines represent the interaction relationship between proteins.

Table IV.

Top five significantly enriched biology processes for differentially expressed genes in three subnetworks.

GO ID	Term	Gene counts	P-value	FDR
Subnetwork 1
GO:0042127	Regulation of cell proliferation	30	3.15×10−20	5.31×10−17
GO:0010647	Positive regulation of cell communication	21	7.11×10−18	1.20×10−14
GO:0009967	Positive regulation of signal transduction	20	1.89×10−17	3.19×10−14
GO:0008284	Positive regulation of cell proliferation	22	3.59×10−17	6.06×10−14
GO:0009719	Response to endogenous stimulus	21	4.01×10−16	7.44×10−13
Subnetwork 2
GO:0022402	Cell cycle process	24	1.05×10−20	1.65×10−17
GO:0007049	Cell cycle	26	3.03×10−20	4.79×10−17
GO:0051726	Regulation of cell cycle	19	3.49×10−18	5.51×10−15
GO:0022403	Cell cycle phase	20	7.57×10−18	1.20×10−14
GO:0051329	Interphase of mitotic cell cycle	13	3.33×10−16	5.22×10−13
Subnetwork 3
GO:0022402	Cell cycle process	23	1.66×10−20	2.59×10−17
GO:0007049	Cell cycle	25	2.88×10−20	4.51×10−17
GO:0051726	Regulation of cell cycle	19	4.22×10−19	6.61×10−16
GO:0022403	Cell cycle phase	19	2.27×10−17	3.55×10−14
GO:0006259	DNA metabolic process	20	3.35×10−17	5.25×10−14

GO, gene ontology; Gene counts, number of genes; FDR, false discovery rate.

Table V.

Top five significantly enriched pathways for differentially expressed genes in three subnetworks.

KEGG ID	Term	Gene counts	P-value	FDR
Subnetwork 1
hsa05200	Pathways in cancer	32	3.04×10−25	3.30×10−22
hsa05219	Bladder cancer	15	4.66×10−19	5.05×10−16
hsa05215	Prostate cancer	14	1.23×10−12	1.34×10−9
hsa05213	Endometrial cancer	11	3.80×10−11	4.12×10−8
hsa05218	Melanoma	12	4.29×10−11	4.65×10−8
Subnetwork 2
hsa04110	Cell cycle	19	3.60×10−20	3.72×10−17
hsa05200	Pathways in cancer	21	5.11×10−15	5.27×10−12
hsa05215	Prostate cancer	11	3.82×10−10	3.94×10−7
hsa04115	p53 signaling pathway	10	7.34×10−10	7.57×10−7
hsa05220	Chronic myeloid leukemia	10	1.81×10−9	1.86×10−6
Subnetwork 3
hsa04110	Cell cycle	18	2.32×10−19	2.37×10−16
hsa05200	Pathways in cancer	19	2.03×10−13	2.07×10−10
hsa04115	p53 signaling pathway	9	9.25×10−9	9.43×10−6
hsa05220	Chronic myeloid leukemia	9	2.04×10−8	2.08×10−5
hsa05223	Non-small cell lung cancer	8	4.40×10−8	4.49×10−5

KEGG, Kyoto Encyclopedia of Genes and Genomes; Gene counts, number of genes; FDR, false discovery rate.

Discussion

TNBC is one of the most deadly breast cancer subtypes due to the lack of an effective treatment. The improvement of the diagnosis and therapeutic methods of this disease relies on the discovery of novel potential molecular markers. In the present study, a total of 121 DEGs between TNBC and normal tissues were identified, consisting of 101 upregulated and 20 downregulated genes. With 121 gene signatures mapped from the STRING database, a giant component PPI network was established with 118 nodes and 1,264 edges. After applying the ClusterONE clustering algorithm, three significant subnetworks with highly connected nodes were obtained. The top five ranked genes as hub nodes with maximum degrees were identified, including TP53, GAPDH, CCND1, HRAS and PCNA. In addition, it was revealed that these hub nodes existed in all of the three subnetworks, which contributed to the biological processes of cell growth and the cell cycle and connected the cell cycle, p53 signaling pathway and pathways in cancer.

TP53, also termed p53, encodes a tumor suppressor protein that controls the cell cycle arrest, apoptosis, senescence, DNA repair and changes in metabolism (19). Mutations in this gene are associated with a variety of human cancer types, including TNBC, lung cancer and high-grade serous ovarian tumor types (20). p53 is mutated and overexpressed in ~25–30% of human breast cancer (21), and had an increased incidence in TNBC (22,23). Nishimura et al (24) also found that high proliferation rate and frequent p53 overexpression occurred in TNBC. Furthermore, p53 is also expressed as smaller isoforms, some of which inhibit wild-type p53. Avery-Kiejda et al (25) reported that Δ40p53, one of the p53 isoforms, was significantly upregulated in tumor tissue when compared with the normal breast and was closely associated with TNBC. Mutation and isoforms of p53 may provide an alternate explanation for the malfunction of p53 pathway and deregulated p53 signaling in TNBC. The present study demonstrated that the p53 gene was elevated in the TNBC samples and that p53 was a hub protein with a degree score of 86 in the established PPI network. Therefore, the p53 gene may be a key regulator in TNBC development.

GAPDH, originally identified as a glycolytic enzyme is considered a housekeeping gene. It is frequently used as an internal standard for gene expression in RNA or protein experiments. However, the abnormal expression of GAPDH has been confirmed to have a close association with various types of cancer (26), as it serves an important role in carcinogenesis and cell death, as well as energy metabolism (27). Notably, increased levels of GAPDH were observed in most types of human cancer and were often correlated with reduced survival (28,29). Results of the present study indicated that GAPDH is predicated as a pivotal gene associated with TNBC and the cell cycle, and may be a potential biomarker for detection and prevention of TNBC.

CCND1, a member of the cyclin-dependent kinase regulator family, is required for the activation of CDK4 and CDK6, and is recognized as a critical modulator of the cell cycle and a positive regulator of cell proliferation (30). Aberrant amplification and overexpression of CCND1 are a driving force in 13–20% of human breast cancer, and are associated with poor disease outcome (31). However, Mylona et al (32) suggested that CCND1 overexpression may serve as a marker for prolonged survival of patients with TNBC. Although the exact role of CCND1 in TNBC remains controversial, according to the findings of the present study, CCND1 overexpression was detected in TNBC tissues, and it acted as a hub gene in the PPI network of TNBC. These data implied that CCND1 is of clinical importance for TNBC.

HRAS, a member of the ras superfamily of genes, encodes for a 21-kDa protein (p21), which takes part in the signal transduction pathways that control proliferation and apoptosis, and regulate the cell cycle. Ras genes are involved in a wide variety of human tumor types and there is a known positive correlation between HRAS activation and breast cancer (33). In addition, previous studies (34,35) have shown that HRAS can induce invasion and migration of the breast cancer cell line, MCF10A. From the present study, HRAS may serve a significant role in the pathogenesis of TNBC and, given its functional significance in various types of cancer, it may also be a potential therapeutic target in the treatment of TNBC.

PCNA is well known as a coordinator of essential cellular processes for cell growth, death, and maintenance. Accumulated evidence suggests that an enhanced level of PCNA often correlates with carcinogenesis and that PCNA levels can be used as a prognostic marker in certain cases (36–39). Yu et al (40) demonstrated that when PCNA activity was inhibited by a peptide, it suppressed the growth of TNBC cells. Given the critical function of PCNA in cancer growth, it is possible that the targeting of PCNA may be a viable therapeutic method for TNBC.

Pathways in cancer and pathways involving the cell cycle were demonstrated to be highly enriched in the DEGs for the PPI network and the subnetworks. All these pathways have been implicated in the development of breast cancer. Notably, the most significantly enriched pathways were pathways in cancer, which refers to some classical cancer pathways, including p53 signaling, Wnt signaling, and Janus kinase-signal transducer and activator of transcription (JAK-STAT) signaling pathways. Certain previous studies (41–43) suggested that these critical signaling nodes, including p53, β-catenin and JAK/STAT3, can be used as therapeutic targets for the treatment of TNBC. In addition, cancer has been viewed as a cell cycle disease (44). Substantial evidence has indicated that the hub genes, P53, CCND1, HRAS and PCNA, are related to the pathways of the cell cycle, which is in accordance with the functions of networks identified in the present study.

In conclusion, 121 genes were revealed to be differentially expressed between TNBC and normal tissues by integrated analysis. Among them were TP53, GAPDH, CCND1, HRAS and PCNA, and these may be involved in TNBC progression via the cell cycle pathways and pathways in cancer. These findings may improve the understanding of the pathogenesis of TNBC and the development of targeted treatments for TNBC. Further experiments are required to confirm the findings of this work and the hypotheses put forward.