Bioinformatics analysis revealing prognostic significance of RRM2 gene in breast cancer

Breast cancer is the most frequently diagnosed tumor and a leading cause of cancer-related deaths among women worldwide [1]. Early diagnosis and treatment strategies including surgery, chemotherapy, radiotherapy, endocrine agents, and biological targeting agents have reduced patient morbidity and mortality; however, the prognosis of breast cancer remains poor. While clinical, pathological, and molecular features are widely used for establishing prognostics and predicting outcomes, finding more sensitive and specific biomarkers as surrogates of these features is the current trend in breast cancer research [2].

Ribonucleotide reductase M2 subunit (RRM2), a rate-limiting enzyme for DNA synthesis and repair, displays vital roles in many critical cellular processes such as cell proliferation, invasiveness, migration, angiogenesis, and senescence [3]. RRM2 is frequently overexpressed in various malignancies and functions like a tumor driver [4–8]. Accumulating evidence has suggested that targeting RRM2 may be a novel strategy for cancer treatment. For example, RRM2 protected glioblastoma cells from endogenous replication stress, DNA damage, and apoptosis; RRM2 inhibition sensitized glioblastoma cells to agent treatment [9]. Knockdown of RRM2 attenuated melanoma growth both in vitro and in vivo, which correlated with maintenance of senescence-associated cell-cycle arrest [10]. In terms of breast cancer, both genetic suppression by RNA interference approach and pharmacological inhibition by small molecular antagonist of RRM2 gene significantly reversed tamoxifen-resistant cell proliferation, reduced cell motility, activated pro-apoptotic pathways, and decreased tumor growth [11–13]. Moreover, it was reported that RRM2 was associated with chemoresistance of breast cancer cells to adriamycin; suppression of RRM2 synthesis could enhance the chemosensitivity to toxic insult [14]. Taken together, these findings suggest that RRM2 may act not only as an oncogene, but also as a promising prognostic biomarker and potential therapeutic target in cancer.

Therefore, in the present study, we evaluated the significance of RRM2 gene expression in breast cancer by using bioinformatics analysis of the clinical parameters and survival data in several large online databases.

The Oncomine (http://www.oncomine.org), an online database containing microarray expression data from a variety of human cancers, was used to determine the level of RRM2 in breast cancer patients and normal individuals with the threshold of fold change ≥ 2, P-value ≤ 1E-4, and gene rank ≥ top 10% [15]. Gene co-expressed with RRM2 was analyzed and displayed as a heat map.

The Breast Cancer Gene-Expression Miner v4.1 (bcGenExMiner v4.1, http://bcgenex.centregauducheau.fr/BC-GEM), a mining tool of published annotated genomics data, was utilized to evaluate the association between RRM2 gene and clinical parameters, as well as the relevance with metastatic relapse event [16,17]. The correlation between RRM2 and KIF11 were generated using the correlation module.

The PrognoScan (http://www.prognoscan.org/) is a large database with clinical annotation and a web-based tool for assessing the biological relationship between gene expression and prognostic information including overall survival, relapse-free survival, distant metastasis-free survival, disease-specific survival, and disease-free survival in breast cancer patients [18]. Cox P-values and hazard ratio (HR) with 95% confidence intervals were calculated automatically.

The Kaplan–Meier Plotter (http://kmplot.com/analysis/), a platform containing gene expression information and survival data of 5143 clinical breast cancer patients, was applied to verify the prognostic value of RRM2 gene in overall survival, relapse-free survival, and distant metastasis-free survival [19].

The heat map of RRM2 and KIF11 in the same patient cohort were constructed by data mining in TCGA Breast Cancer using the UCSC Xena browser (http://xena.ucsc.edu/).

We first checked the expression of RRM2 gene in 20 types of malignant tumor using the Oncomine database. Increased level of RRM2 (red) was observed in gastrointestinal cancers, gynecological cancers, urogenital cancers, and breast cancer (Figure 1). Our analysis also revealed that RRM2 was significantly higher expressed in male breast carcinoma, intraductal cribriform breast adenocarcinoma, invasive breast carcinoma, invasive lobular breast carcinoma, invasive ductal breast carcinoma, ductal breast carcinoma in situ, invasive ductal breast carcinoma epithelia, and ductal breast carcinoma, compared with the corresponding normal tissues (Figure 2A–H and Table 1).

By using the bc-GenExMiner online tool, we next compared RRM2 expression among groups of patients, according to different clinical parameters. Regarding age, there was no significant difference between ≤51- and >51-year group (Figure 3A and Table 2). The Scarff–Bloom–Richardson (SBR) is a histological grade that evaluates tubule formation, nuclear characteristics of pleiomorphism, and mitotic index. The Nottingham Prognostic Index (NPI) has been validated to stratify patients into additional prognostic groups according to tumor size, lymph node stage, and tumor grade. Breast cancer patients with more advanced SBR grade and NPI tended to express higher RRM2 gene (Figure 3B,C). Estrogen receptor (ER) and progesterone receptor (PR) status were negatively associated with RRM2 expression (Figure 3D,E and Table 2). Conversely, human epidermal growth factor receptor-2 (HER-2) status was confirmed to correlate positively with RRM2 expression (Figure 3F and Table 2). Breast cancer patients with positive nodal status (N) showed increased level of RRM2 than those with negative nodal status (Figure 3G and Table 2). Besides, we found that RRM2 was strongly elevated in basal-like subtype with respect to non-basal-like subtype; the same pattern of change was also observed in triple-negative breast cancer (TNBC) patients (Figure 3H,I and Table 2).

We then investigated the prognostic value of RRM2 gene. The Kaplan–Meier curves indicated that lower level of RRM2 correlated with preferable overall survival (Figure 4A). While breast cancer patients with up-regulated RRM2 demonstrated worse relapse-free survival (Figure 4B), patients with decreased RRM2 expression presented better distant metastasis-free survival (Figure 4C). Furthermore, RRM2 exerted positive effect on metastatic relapse event, as suggested by the forest plot using the bc-GenExMiner tool (Figure 4D). The PrognoScan database showed that overexpression of RRM2 was significantly associated with inferior overall survival, relapse-free survival, distant metastasis-free survival, disease-specific survival, and disease-free survival (Table 3).

To further investigate the underlying regulation of RRM2 in breast cancer, data mining of the co-expression of RRM2 gene was performed using the Oncomine database. The co-expression profile of RRM2 was identified with a large cluster of 1802 genes across 61 breast carcinomas, and KIF11 is a principal correlated gene (Figure 5A). Data mining in bc-GenExMiner revealed a positive correlation between RRM2 and KIF11 (Figure 5B). By comparing the RRM2 and KIF11 expression heat map derived from the UCSC Xena web-based tool, it was confirmed that RRM2 expression gradually elevated with increasing KIF11 transcript level, which was determined among a 50-gene qPCR assay (PAM50) breast cancer subtypes in TCGA database (Figure 5C). These data indicated that RRM2 could be associated with the KIF11 signaling pathways in breast cancer.

RRM2 plays vital roles in diverse cellular functions such as cell proliferation, invasiveness, migration, angiogenesis, senescence, and tumorigenesis [3]. It was reported that RRM2 was associated with resistance of breast cancer cells to chemotherapy and endocrine agents and that targeting RRM2 may be a novel strategy for cancer treatment [11–14]. However, the significance of RRM2 expression in prognosis of breast cancer remains largely unclear.

In the present work, we analyzed the expression profile of RRM2 by Oncomine database. RRM2 gene was higher expressed in male breast carcinoma, intraductal cribriform breast adenocarcinoma, invasive breast carcinoma, invasive lobular breast carcinoma, invasive ductal breast carcinoma, ductal breast carcinoma in situ, invasive ductal breast carcinoma epithelia, and ductal breast carcinoma patients, with respect to normal individuals. By using the bc-GenExMiner online tool, we found that ER and PR were negatively correlated with RRM2 expression. Conversely, SBR, NPI, HER-2 status, nodal status, basal-like status, and triple-negative status were positively related to RRM2 level in breast cancer samples with respect to normal tissues. As known to all, patients with ER or PR negative, nodal positive, HER-2 positive, basal-like or triple-negative status generally display an unsatisfied therapeutic response and worse clinical outcome. Therefore, our results suggested that lower expression of RRM2 may predict a better prognosis of breast cancer.

We further investigated the prognostic value of RRM2 in breast cancer using the Kaplan–Meier Plotter, PrognoScan, and bc-GenExMiner databases. Patients with increased RRM2 showed worse overall survival, relapse-free survival, distant metastasis-free survival, disease-specific survival, and disease-free survival. Additionally, high RRM2 expression was correlated with an increased risk of metastatic relapse event, as suggested by the forest plot. These findings collectively demonstrated that the level of RRM2 might be a useful predictive biomarker for prognosis of breast cancer. We finally analyzed the co-expression of RRM2 using the Oncomine, bc-GenExMiner, and UCSC Xena web-based tools and confirmed that KIF11 gene was positively correlated with RRM2 expression. KIF11, a molecular motor protein involved in mitosis, was critical for proliferation and self-renewal in chemoresistant breast cancer cells [20]. KIF11 knockdown inhibited tumor growth both in vitro and in vivo, and its expression was responsible for shorter survival time [21]. Thus, our data indicated that RRM2 might contribute to breast cancer progression and drug insensitivity associated with KIF11 expression.

In summary, the present bioinformatics analysis showed that RRM2 was overexpressed in breast cancer patients with respect to normal tissues and was associated with a worse survival. RRM2 could be used as a predictive biomarker for prognosis of breast cancer with co-expressed KIF11 gene. Further studies are needed to more precisely elucidate the value of RRM2 in evaluating breast cancer prognosis.

This work was supported by the Natural Science Foundation of China [grant number 81702591]; and the Natural Science Foundation of Jiangsu Province [grant number BK20170294].

The authors declare that there are no competing interests associated with the manuscript.

Conceived and designed the experiments: W.-X.C. and L.-G.Y. Analyzed the data: W.-X.C., L.-Y.X., and L.C. Contributed analysis tools: Q.Q., L.S., and Y.-L.Z. Wrote the paper: W.-X.C.

bc-GenExMiner

Breast Cancer Gene-Expression Miner

ER

estrogen receptor

HER-2

human epidermal growth factor receptor-2

NPI

Nottingham Prognostic Index

PR

progesterone receptor

RRM2

ribonucleotide reductase M2 subunit

SBR

Scarff–Bloom–Richardson