Expression Pattern and Prognostic Value of EPHA/EFNA in Breast Cancer by Bioinformatics Analysis: Revealing Its Importance in Chemotherapy

The activities of the ephrin family in breast cancer (BrCa) are complex. Family A receptors (EPHA) and ligands (EFNA) can act as oncogenes or tumor suppressors and are implicated in chemoresistance. Here, we examined the expression pattern and prognostic value of the EPHA/EFNA family in patients with breast cancer, including patients with different subtypes or different chemotherapy cohorts. In the UALCAN database, the mRNA expression of EPHA1, EPHA10, EFNA1, EFNA3, and EFNA4 was significantly higher, whereas that of EPHA2, EPHA4, EPHA5, and EFNA5 was significantly lower in breast cancer tissues than in paracancerous tissues. The transcriptional levels of EPHA/EFNA family members were correlated with intrinsic subclasses of breast cancer. The relationship between EPHA/EFNA and the clinicopathological parameters of BrCa was analyzed using bc-GenExMiner V4.5. EPHA1, EPHA2, EPHA4, EPHA7, EFNA3, EFNA4, and EFNA5 were upregulated in estrogen receptor- (ER-) and progesterone receptor- (PR-) negative tumors, whereas EPHA3, EPHA6, and EFNA1 were upregulated in ER- and PR-positive tumors. EPHA1, EPHA2, EFNA3, and EFNA4 mRNA expression was significantly higher in human epidermal growth factor receptor 2- (HER2-) positive tumors than in HER2-negative tumors. Triple-negative status was positively correlated with EPHA1, EPHA2, EPHA4, EPHA7, EFNA3, EFNA4, and EFNA5 and negatively correlated with EPHA3 and EPHA10 mRNA expression. Genetic alterations of EPHA/EFNA in breast cancer varied from 1.1% to 10% for individual genes, as determined by the cBioPortal database. The Kaplan–Meier plotter indicated that high EphA7 mRNA expression was associated with poor overall survival (OS) and recurrence-free survival (RFS), especially in the HER2 and luminal A subtypes. EFNA4 was predicted to have poor OS and RFS in breast cancers, especially in luminal B, basal-like subtype, and patients treated with adjuvant chemotherapy. High EPHA3 expression was significantly associated with better OS and RFS, especially in the luminal A subtype, but with poor RFS in BrCa patients receiving chemotherapy. Our findings systematically elucidate the expression pattern and prognostic value of the EPHA/EFNA family in BrCa, which might provide potential prognostic factors and novel targets in BrCa patients, including those with different subtypes or treated with chemotherapy.


Introduction
Breast cancer is the most commonly diagnosed female cancer and the cause of 685,000 cancer mortality in 2020 worldwide [1]. Tumor recurrence and metastasis contribute to the high death rate [2]. Despite extensive research into the treatment of BrCa, chemotherapy resistance is an important issue limiting the efficacy of treatment. Novel biomarkers to predict prognosis or the sensitivity to chemotherapy are urgently needed.
Receptor tyrosine kinases (RTKs) play an important role in a variety of cellular processes in cancer [3]. Ephrins, also known as ephrin ligands, and Eph receptors (Ephs), which are RTKs, are key regulators of physiological and pathological processes involved in development and disease, such as cellular motility, cell repulsion, and cell adhesion [4]. The Ephrin family consists of multiple Ephs and ephrins. Both receptors and ligands are membrane-bound proteins that require direct cell-cell interaction for activation. Eph/ephrin signal transduction occurs not only in the receptorexpressing cell but also in the ligand-expressing cell via bidirectional signaling [5]. The ligands can have a glycosylphosphatidylinositol anchor (A type) or a membrane-spanning protein domain (B type). The receptors are also categorized as A or B according to the type of ligand they bind to. Ephrin family A includes ten receptors named EPHA (1)(2)(3)(4)(5)(6)(7)(8)(9)(10) and five ligands designated as EFNA (1)(2)(3)(4)(5) [6]. The interaction between ligands and receptors via bidirectional signaling and its involvement in cancer biology are mediated by complex processes [7,8]. Several Ephrin A (EPHA/EFNA) family members are overexpressed or downregulated in a variety of tumors, suggesting that they act as oncogenes or as tumor suppressors according to the cellular context [9]. Ephrin A family members that are overexpressed in cancer including EFNA1 in melanoma [10]; EPHA2 in prostate cancer [11], nasopharyngeal carcinoma [12], and squamous-cell carcinoma of the head and neck [13]; and EPHA3 in non-small-cell lung cancer [14]. EPHA7 acts as a tumor suppressor in follicular lymphoma and is a potential therapeutic target [15]. EphA2 expression levels are associated with the invasiveness and aggressive behavior of BrCa [16]. The role of EPHA/EFNA as tumor suppressors in breast carcinogenesis was also demonstrated by targeting EPHA2 [17]. EPHA2 regulates the sensitivity to paclitaxel in nasopharyngeal carcinoma via the phosphoinositide 3-kinase/Akt signaling pathway [18]. Cisplatin chemotherapy-induced ERK1/2-RSK1/2-EphA2-GPRC5A signaling is related to acquired chemoresistance in ovarian cancer [19]. Previous findings indicate that EPHA/EFNA family members may serve as biomarkers for predicting prognosis or the response to treatment, which prompted us to analyze the expression pattern and prognostic role of the EPHA/EFNA family in BrCa.
Online databases provide access to a wealth of information, such as microarray RNA chips from the Gene Expression Omnibus (GEO) [20] and RNA sequences from The Cancer Genome Atlas (TCGA) [21]. In this study, we compared the transcriptional levels of EPHA/EFNA in BrCa and paracancerous tissues using the UALCAN database [22]. The relationship between members of the EPHA/EFNA family and clinicopathologic characteristics of BrCa patients was analyzed using breast cancer gene expression miner (bc-GenExMiner) v4.5 [23,24]. Genetic alterations of EPHA/EFNA family members, including mutations and putative copy number alterations (CNAs), were analyzed in cBioPortal [25]. Moreover, the association of EPHA/EFNA gene expression with clinical outcomes was assessed in patients with BrCa, including patients with different subtypes or those undergoing chemotherapy, using the Kaplan-Meier plotter database [26].

Methods and Materials
2.1. Gene Expression Analysis in UALCAN. UALCAN (http:// ualcan.path.uab.edu/) is an online comprehensive and interactive platform based on RNA-seq data and clinical information from TCGA database [22]. In this study, UALCAN was used to investigate the transcriptional levels of EPHA/EFNA family members in primary BrCa tissues and their association with the subtype.

2.2.
Analysis of the Relationship between EPHA/EFNA and Clinicopathologic Characteristics in Bc-GenExMiner. Bc-GenExMiner v4.5 (http://bcgenex.centregauducheau.fr/BC-GEM), a statistical mining tool of published BrCa transcriptomic data, was used to evaluate the relationship between EPHA/EFNA family members and the clinicopathologic characteristics of BrCa patients, including age, nodal status, estrogen receptor (ER) status, progesterone receptor (PR) status, human epidermal growth factor receptor 2 (HER2) status, lymph node status, triple-negative tumors, and P53 status (sequence). Information on ER, PR, HER2, lymph node status, and pathological grade was not available for all 3,996 patients.

Genetic Alterations Analysis in cBioPortal.
OncoPrint is a feature of cBioPortal, an open-source web application that allows researchers to explore and analyze cancer genomic datasets (http://www.cbioportal.org/) [27,28]. In this study, genetic alterations of EPHA/EFNA family members including mutations and putative CNAs were analyzed. A total of 1084 tumor samples with RNA-seq data on cBioPortal were included in the study.

The Prognostic
Value of EPHA/EFNA mRNA Expression in BrCa Patients including Those with Different Subtypes and Undergoing Chemotherapy. The Kaplan-Meier plotter is an online database that facilitates predicting the effect of gene expression on survival in cancer. Sources for the database include GEO, EGA (European Genome-Phenome Archive), and TCGA. This platform, which contains gene expression information and survival data of BrCa patients, was used to perform a meta-analysis to verify the prognostic value of EPHA/EFNA family members for predicting overall survival (OS) and relapse-free survival (RFS). Additional analyses were restricted to cohorts according to subtype and including the patients treated with chemotherapy. Kaplan-Meier survival plots were used to compare the prognosis of all cohorts. Hazard ratios (HRs), 95% confidence intervals (CIs), and logrank P values were calculated and displayed online.         [29]. Here, we analyzed the transcriptional levels of EPHA/EFNA family members according to BrCa subtype using the UALCAN database. As shown in Figure 2, EPHA3, EPHA5, EPHA6, and EPHA10 mRNA levels were low in HER2-positive and triple-negative BrCa patients, whereas other EPHA family members did not show this trend. EPHA3, EPHA5, EPHA6, and EPHA10 were expressed at high levels in the luminal subtype (Figures 2(c), 2(e), 2(f), and 2(h)). The highest mRNA expression levels of EFNA were detected in triple-negative tissues except EFNA3, whereas the luminal subtype showed the lowest EFNA mRNA levels (Figures 2(i)-2(l)) except EFNA1. Taken together, these findings indicate that the transcriptional levels of EPHA/EFNA family members are correlated with intrinsic subclasses in BrCa patients.
3.4. Genetic Alterations of EPHA/EFNA in BrCa. Different kinds of genetic alterations, such as missense mutations, amplification, and deep deletions, regulate cancer-related gene expression and participate in oncogenesis. We speculated that genetic alterations may regulate the transcriptional levels of EPHA/EFNA. To investigate mutations and CNAs of the EPHA/EFNA family in BrCa, the OncoPrint feature of cBioPortal (http://www.cbioportal.org) was used to investigate the proportion and percentage of specimens with genetic alterations in EPHA/EFNA. The frequency of alterations in these genes among BrCa samples varied from 1.1% to 10% for individual genes as shown in Figure 3.

Discussion
The activities of the EPHA/EFNA family in BrCa are complex and paradoxical. The expression and prognostic value of EPHA/EFNA in BrCa have not been extensively investigated. In the present study, two cancer databases were used to analyze the transcriptional levels of EPHA/EFNA family 14 BioMed Research International members in BrCa and paracancerous tissues, as well as their association with the BrCa subtype and clinicopathological features. The genetic alterations of EPHA/EFNA family members, including mutations and putative CNAs, were analyzed using cBioPortal. The Kaplan-Meier plotter was used to analyze the association between the expression levels of EPHA/EFNA and OS or RFS in BrCa patients, as well as RFS in different BrCa subtypes, including patients undergoing chemotherapy. The results may be valuable in identifying new BrCa biomarkers for predicting prognosis or sensitivity to chemotherapy and suggest that EPHA and EFNA play both oncogenic and tumor suppressor roles in BrCa. Expression analysis showed that the transcriptional levels of EPHA2, EPHA4, and EPHA5 were significantly lower in BrCa tissues than in nontumor tissues and EPHA1 and EPHA10 were significantly upregulated in BrCa tissues. Brantley et al. investigated EPHA/EFNA protein expression in BrCa and found that EPHA2, EPHA4, and EPHA7 were significantly upregulated in BrCa samples relative to normal controls. The discrepancy between the protein and mRNA 15 BioMed Research International levels of EPHA2 may be due to the fact that high levels of EPHA2 in tumor cells are the result of increased protein stability [31]. In certain malignant breast cell models, EPHA2 protein levels are 50-to 500-fold higher despite comparable levels of EphA2 mRNA [32,33]. Discrepancies in EPHA2 expression may also result from the inclusion of both invasive and noninvasive breast tumors in the TCGA database. The mRNA level of EPHA2 is higher in invasive tumors than in normal breast cells; however, EPHA2 expression is lower in noninvasive breast tumors than in normal breast cells [34]. In this study, EFNA was upregulated in BrCa tissues except for EFNA5. EPHA2 is the dominant and the best characterized EPHA receptor in the BrCa. The role of EPHA2 in breast tumor progression is controversial, and conflicting data on the clinical significance of EPHA2 have been reported in different studies [35,36]. For example, data show that EPHA2 is overexpressed in BrCa clinical samples; however, there is also evidence that EPHA2 acts as a tumor suppressor in breast carcinogenesis [17]. The malignant behavior of EPHA2 is mediated by ligand-independent signaling, and its antioncogenic properties are attributed to ligand-dependent signaling [37]. The crosstalk between EphA2 and BrCa oncogenic pathways promotes tumor cell malignancy in ligand-independent signaling [38,39]. We found an inverse correlation between EPHA2 and the mRNA expression of A-type ligands in the database, as shown by the downregulation of EPHA2 and EFNA overexpression in BrCa cells. Ligand upregulation in the tumor indicates that the ligand-dependent pathway is dominant in the database. Downregulation of the EPHA2 receptor by the ephrin ligand may involve ligand-mediated receptor internalization. In general, EPHA2 negatively regulates tumor growth and migration after canonical ligand-induced EPHA2 signaling, which inhibits the AKT-mTORC1 and MAPK pathways [17]. The interaction between EPHA2 and its ligand activates a negative feedback pathway mediated by growth factoractivated RAS signaling [40]. EFNA1 is upregulated in noninvasive breast cells, thereby inhibiting invasiveness, whereas EFNA1 is downregulated in invasive tumors, allowing EPHA2 to participate in invasion [34].  We also compared the differential transcriptional levels of EPHA/EFNA family members according to the different intrinsic subtypes of BrCa. The results showed differences in the expression patterns between the BrCa subtypes. EPHA3, EPHA5, EPHA6, and EPHA10 were expressed at the highest levels in luminal tissues, whereas HER2-positive and triple-negative tissues tended to express lower levels of EPHA3, EPHA5, EPHA6, and EPHA10. The highest mRNA expression levels of EFNA were found in triple-negative tissues except EFNA3, whereas the lowest mRNA expression levels of EFNA were found in the luminal subtype except EFNA1. Analysis of the bc-GenExMiner database showed that EPHA1, EPHA2, EPHA4, EPHA7, EFNA3, EFNA4, and EFNA5 were upregulated in ER-and PR-negative patients and positively correlated with triple-negative status. Consistent with this, some studies have shown that EPHA2 overexpression in BrCa is negatively correlated with ER and PR status [32,41,42]. EPHA2 overexpression decreases estrogen dependence and tamoxifen sensitivity [43], and EPHA2 is preferentially expressed in the basal-like phenotype [44]. In contrast, EPHA2 expression is also negatively regulated by ERα [41] and wild-type p53 [45]. The correlation of EPHA2 with HER2-positive status in the present study is consistent with the results of previous studies [39,46], and EPHA2 is associated with resistance to trastuzumab therapy [47].
Analysis of the other EPHA/EFNA family members showed that EPHA1 was significantly upregulated in BrCa tissues compared to paracancerous tissues. A previous report showed that EPHA1 downregulation is associated with the invasiveness of breast carcinoma cells [34]. Consistent with the present profiling analyses, EPHA4 expression is associated with basal-like BrCa [48] and EPHA5 has been reported to act as a tumor suppressor, which may be related to abrrant promoter methylation [49]. Liu et al. [50] showed that EPHA7 mRNA was downregulated in BrCa specimens and loss of EPHA7 expression is more common in high-grade, early TNM-stage patients, without lymph node metastasis and correlation with negative HER2 status. Correspondingly, the present data also found downregulation of EPHA7 expression in BrCa specimens. EPHA10 is the only kinasedeficient Eph receptor [51]. The present data indicate that  EPHA10 is upregulated in BrCa tissues, which is similar with a previous study showing that EPHA10 expression is significantly lower in invasive than in noninvasive breast tumors, and is absent in normal cells [34]. Similar with the previous study, the present data showed EFNA4 was upregulated in TNBC [52]. EFNA4 is required for proper differentiation and polarization of mammary epithelial cells, signifying a biologic basis for the overexpression of EFNA4 in BrCa [53].
The most common cancer-related genetic alteration is the DNA CNAs. The OncoPrint feature of cBioPortal was used to determine the frequency of genetic alterations in the EPHA/EFNA family. The results showed that EPHA/EFNA was not frequently amplified. This finding suggests that the EPHA/EFNA family does not affect BrCa survival through DNA alterations, whereas it may affect BrCa through alterations of the interaction network. The copygain frequency of EFNA1, EFNA3, and EFNA4 accounts for a large proportion of BrCa samples, which may explain the significant upregulation of EFNA1, EFNA3, and EFNA4 in BrCa tissues from UALCAN.
The Kaplan-Meier plotter is a survey of public microarray data repositories of survival from 3955 patients with BrCa. Besides, in general patients, we analyzed the prognostic value of EPHA/EFNA mRNA in different BrCa subtypes and in BrCa patients treated with chemotherapy. Large-scale expression profiling studies revealed a negative association between the overexpression of EPHA2, EPHA4, and EPHA7 and overall and disease-free survival in BrCa [54]. Consistent with this report and its positive correlation with the triplenegative (TNBC) status in the bc-GenExMiner, the Kaplan-Meier plotter indicated that high EPHA7 mRNA expression predicted a poor OS and RFS in BrCa, especially in the HER2+ and luminal A subtypes. EFNA4 was associated with poor OS and RFS in BrCa, which is consistent with its positive correlation with the TNBC status. In particular, EFNA4 was associated with poor RFS in luminal B and basal-like subtypes and in BrCa patients treated with adjuvant chemotherapy. High mRNA expression of EPHA3 was significantly associated with better OS and RFS, especially in the luminal A subtype, which is consistent with its negative correlation with the TNBC status. However, in patients receiving adjuvant chemotherapy and neoadjuvant chemotherapy, EPHA3 is a risk factor. These results suggest that EPHA3, EPHA7 and EFNA4 are involved in BrCa development and could predict the prognosis of patients with BrCa in various subtypes and in those receiving different chemotherapy regimens. BrCa is a heterogeneous disease with subtypedependent histopathological and clinical significance. In the subgroup analysis, the triple-negative (basal-like) subtype is a unique subtype of BrCa with a poor prognosis and more likely to develop chemoresistance [55]. In this study, we showed that low expression of EPHA1, EPHA4, EPHA5, and EPHA7 and high expression of EFNA1, EFNA4, and EFNA5 predicted an unfavorable prognosis in basal-like patients. Although we found that high mRNA expression of EPHA2 was associated with a longer RFS in patients with BrCa, it was a risk factor in the cohort receiving adjuvant chemotherapy. In fact, the RSK1/2-EphA2-GPRC5A oncogenic signaling association with platinum chemotherapy resistance has been reported in recent study [19] . Furthermore, in clinical practice, these biomarkers coupled with the specific role of EPHA/EFNA signaling in cancer could promote targeted therapeutics, as reported before [52,56].
The present study has several limitations that should be addressed in future studies. First, mRNA levels are not always indicative of a functional protein and thus may not fully represent the protein expression of EPHA/EFNA [25]. Future studies should include protein detection techniques to accurately assess the protein levels. Second, the function of EPHA3 and EFNA7 in BrCa has not been studied in BrCa and needs to be investigated in the future. Third, multivariate analysis could not be used in the database to correct the associations between different clinicopathological features.
In conclusion, we showed that the EPHA/EFNA family is widely expressed in BrCa tumor cells. EPHA/EFNA were identified as prognostic factors and potential targets for BrCa, which may improve our understanding of the complexity and heterogeneity of BrCa at the molecular level. The present results may help develop tools to accurately predict prognosis and design customized therapies.