Prominin-1 (PROM1) Signicantly Correlated With Bone Metastasis and Inuenced Prognosis in Breast Cancer

Background This study aimed to explore the important biomarker associated with bone metastasis (BM) in breast cancer (BRCA). Methods The GSE175692 dataset was used to detect signicant differential expressed genes (DEGs) between BRCA samples with or without BM, and important pathways were then explored. Further, we constructed protein-protein interaction (PPI) network on GEGs and ltered 5 vital nodes. Through performing cox regression, Kaplan-Meier analysis, nomogram, ROC curve, and risk score model, signicant prognostic factor was gradually identied. Finally, gene set enrichment analysis (GSEA) analysis was performed to reveal the potential mechanism.


Introduction
Metastasis accounted for more than 90% of cancer-related mortality [1]. Bone is one of the most preferred sites of metastatic spread from different cancer types, including breast cancer (BRCA). BRCA has become a critical health care issue that substantially affected women worldwide. Besides lung, liver and brain, the most common site for metastasis in BRCA is the bone [2]. Bone is also the rst site of distant metastasis in 25% to 40% of patients with advanced BRCA [3]. It should be noted that different subtypes of BRCA exhibited distinct metastatic behavior in terms of kinetics and anatomic sites of relapse [4,5]. For example, bone-only metastases were more common in the HR-positive group than in the other subtypes Page 3/24 [6]. The bone metastasis in BRCA was able to in uence the survival of patients. It was reported that 5-year survival rate of BRCA patients without metastasis was greater than 95% but close to only 20% once bone metastasis occurred [7]. In addition, BRCA bone metastasis also caused series of bone-related complications such as pain, pathologic fractures and spinal cord compression, which signi cantly affected the patient's quality of life [6], also increased the medical costs and mortality risk. As such, new targets and therapeutic strategies associated with BRCA bone metastasis are urgently required.
Over the past few decades, a great deal of biomarkers about BRCA bone metastasis has been generated, which facilitated the studies on cancer etiology. Pantano et al. found that integrin alpha5 (ITGA5) was highly expressed in bone metastases, compared to lung, liver, or brain metastases [8]. Awolaran et al. identi ed 15 proteins expressed by BRCA cells as factors that mediated BRCA bone metastasis, and upregulation of them could promote BRCA metastasis to bone, except for C-C Motif Chemokine Ligand 2 (CCL2) showed a reduced expression [7]. Westbrook et al. identi ed and validated the dedicator of cytokinesis protein 4 (DOCK4) as a potential biomarker for risk of bone metastasis development in patients with early BRCA [9]. In addition, Zhang et al. found that microRNA-429 can inhibit BRCA bone metastasis by regulating matrix metallopeptidase 9 (MMP-9) [10]. More and more useful biomarkers should be detected to reveal the cancer etiology involved in BRCA bone metastasis.
The present study was conducted to detect signi cant bone metastasis-associated biomarkers involved in BRCA. The differential expressed genes (DEGs) between BRCA bone metastasis and non-metastasis samples were initially screened out through the GEO dataset. Then, the impacts of signi cant DEGs on the survival probability of metastasis BRCA samples were predicted. Regarding signi cant prognostic biomarkers, we explored associated regulatory pathways involved in BRCA metastasis. This study was conducive to reveal useful biomarkers and potential mechanisms involved in BRCA bone metastasis.

DEGs identi cation and function analysis
GSE175692 dataset was used to detect signi cant differential expressed genes (DEGs), which contained 33 BRCA bone metastasis samples and 151 non-bone metastasis samples. According to the threshold of P<0.05 and absolute log fold change (FC)>1, the DEGs between the 2 groups were identi ed by the limma R package. And top 20 up and down-regulated DEGs were presented. Subsequently, the signi cant KEGG pathway and GO terms about DEGs were explored through clusterPro ler R package. GO terms were annotated through the biological process (BP), cellular component (CC) and molecular function (MF) aspects.

PPI network construction and hub genes determination
The protein-protein interaction (PPI) network among DEGs was constructed using the online tool String with setting interacting score as medium, followed by visualization through Cytoscape. The top 10 hub genes among the whole PPI network were detected by the Degree gene ranking method using Cytohubba plug-in of Cytoscape. Venn analysis was then performed to lter more important nodes between the top 10 hub genes and top20 DEGs. Finally, 5 signi cant nodes were identi ed for further investigation.

Expression analysis on hub genes
The mRNA differential expression of 5 hub genes between BRCA samples with or without bone metastasis was rstly evaluated using the dataset of GSE175692. Then, the GSE14020 dataset was used to verify their mRNA expression, which contained 18 bone metastasis and 47 non-bone metastasis BRCA samples. We also investigated their mRNA expression between normal and BRCA samples based on TCGA data in the GSCA database. In addition, protein expression of hub genes in BRCA was also evaluated based on CPTAC data in UALCAN.

Prognosis analysis on hub genes
The impact of 5 hub genes on the survival probability of BRCA metastasis samples was initially evaluated using the GSE175692 dataset through univariate cox regression analysis. Kaplan-Meier analysis was performed to explore the in uence of hub genes on the overall survival time of BRCA metastasis samples. Further, nomogram and receiver operating characteristic (ROC) curve analyses were conducted to reveal the prognostic performance of hub genes. A risk score model contained hub genes were also constructed to assess the potential prediction ability. Moreover, the prognostic impact of hub genes on the overall survival of BRCA patients was also veri ed using the GSE124647 dataset, which included 140 BRCA metastasis samples. Regarding the signi cant prognosis factor, its prognostic impact on distant metastasis-free survival of BRCA patients was veri ed as well using the GSE11121 dataset, which contained 200 BRCA samples.

GSEA analysis on signi cant prognostic factors
Due to the differential expression and vital prognostic effect of signi cant prognostic factors in metastasis BRCA samples, we performed the gene set enrichment analysis (GSEA) analysis to explore the possible mechanism associated with BRCA metastasis. The gene expression pro le for GSEA analysis was obtained from the GSE175692 dataset, which included 184 BRCA metastasis samples. All the patients were divided into high and low expression groups according to the median gene expression level. Then, the potential pathway enriched in high/low expression groups was predicted by GSEA analysis. The threshold for GSEA analysis was set as follows: the number of permutations (1000), enrichment statistic (weighted), metric for ranking genes (Pearson).

Statistical analysis
The expression difference between the 2 groups was compared with independent samples T-test or Mann-Whitney test, and results were presented with Violin-chart and Box-scatter which contained the expression median, upper quartile, lower quartile, maximum and minimum. The effects of gene expression on the survival of patients were evaluated through survival analysis and Cox regression showing hazard ratio (HR) and 95% con dence interval (CI). P<0.05 was considered as the statistical signi cance.

DEGs identi cation and function analysis
The GSE175692 dataset was used to lter the DEGs between BRCA samples with or without bone metastasis. A total of 74 signi cant DEGs were found according to absolute LogFC>1 and P-value<0.05, and the top 10 up-regulated and 10 down-regulated DEGs were marked (Fig.1A). KEGG analysis showed that 74 DEGs were mainly associated with infectious disease, signaling molecules and interaction (Fig.1B). The enriched pathway of partial DEGs was shown in Fig.1C.
Functional enrichment analysis was then performed to reveal the role of 74 DEGs in cancer progression ( Fig.2). The signi cant enriched KEGG pathway included ECM-receptor interaction, Cytokine-cytokine receptor interaction and Human papillomavirus infection. For cellular component, the DEGs were largely located at the collagen-containing extracellular matrix and extracellular matrix. For biological process, DEGs primarily participated in extracellular structure organization. In terms of molecular function, DEGs were signi cantly enriched in receptor regulator activity and receptor ligand activity.

PPI network and signi cant hub genes analysis
The protein-protein interaction (PPI) network among DEGs was constructed using String and Cytoscape. The PPI network contained 60 nodes and 167 edges (Fig.3). Using the Degree gene ranking method, we The consistent genes between the top 20 DEGs and 10 hub genes were determined to lter more signi cant hub genes. The 5 most signi cant hub genes were subsequently found which contained 3 upregulated (MMP9, ITGB3 and BMP2) and 2 down-regulated (CCL2 and PROM1) DEGs (Fig.4A). The detailed information of 5 hub genes in GSE175692 was presented in Fig.4B. Differential expression of 5 hub genes between bone metastasis and none bone metastasis BRCA samples were then compared, and results of quantity analysis were shown in Fig.4C. Differential expression of 5 hub genes between BRCA bone metastasis and non-bone metastasis samples was also veri ed using the GSE14020 dataset. The result showed that the comparison result of MMP9, ITGB3, BMP2 and PROM1 in GSE14020 conformed to that in GSE175692 (Fig.5). However, no difference was observed in terms of CCL2 mRNA expression in GSE14020.
We then explored the expression of 5 hub genes between normal and BRCA samples. The results showed that mRNA expression of PROM1, CCL2 and BMP2 were down-regulated in BRCA samples compared with that in the normal group (Fig.6A, all FDR<0.05). No difference was observed between normal and BRCA groups in terms of ITGB3 and MMP9 mRNA expression. The protein expression of MMP9, ITGB3 and PROM1 were all downregulated in BRCA compared with that in the normal group (Fig.6B, all P<0.05). It should be noted that CCL2 and BMP2 were not identi ed in Breast data set.

Prognosis analysis on hub genes
We rstly performed the univariate cox regression to screen the prognosis-related biomarkers in metastatic BRCA and found that PROM1 and CCL2 showed a signi cant correlation with the overall survival of patients (Fig.7A). Further survival analysis indicated that high expression of PROM1 and CCL2 shortened the overall survival time of metastatic BRCA patients (Fig.7B).
Subsequently, the prognostic performance of PROM1 and CCL2 in metastatic BRCA was evaluated. A nomogram analysis showed that PROM1 possessed the largest contribution to the survival probability of patients, contributing 100 points (Fig.8A). The ROC analysis indicated that the prediction ability of PROM1 was slightly superior to CCL2 (Fig.8B).
According to the gene expression levels and clinical outcome, all the patients with metastatic BRCA were divided into high-risk and low-risk groups. The detailed information for risk score was presented in Fig.9A. For each individual, the risk score=0.076806616×(expression of PROM1) 0.12554159×(expression of CCL2). Further, we found that the high-risk group showed a poor prognosis than the low-risk group (Fig.9B). The ROC curve presented the prediction ability of risk score on the survival probability (Fig.9C).
The nomogram indicated that risk score possessed the best prognostic performance than PROM1 and CCL2 (Fig.9D). These results showed that the risk score model possessed favorable prediction ability on the survival probability of metastatic BRCA patients.
The above results indicated the signi cance of PROM1 and CCL2 in metastatic BRCA, we further veri ed their prognostic impacts using the GSE124647 dataset, which contained 140 samples with metastatic BRCA. The survival analysis in Fig.10 showed that PROM1 still presented a signi cant prognostic impact. However, we found no impact of CCL2 on the survival time of metastatic BRCA patients.
The PROM1 and CCL2 exerted a signi cant effect on the survival of metastatic BRCA patients, we also explored their prognostic impacts on the survival of BRCA patients with single bone metastasis. However, both PROM1 and CCL2 exerted no signi cant impacts on the survival of BRCA patients with single bone metastasis (Fig.11), which might be due to the insu cient sample size.
We also used the GSE11121 dataset to verify the prognostic impact of PROM1 on the distant metastasisfree survival of BRCA samples. The best cutoff point of PROM1 expression for survival analysis was presented in Fig.12A. Survival analysis showed that PROM1 high expression was not conducive to the distant metastasis-free survival of BRCA patients (Fig.12B, P<0.001), indicating the correlation of PROM1 expression with distant metastasis in BRCA.

GSEA analysis on PROM1
PROM1 was determined as a vital down-regulated gene in bone metastasis and played an important role in metastasis BRCA progression. Finally, we explored the potential mechanism of PROM1 involved in metastasis BRCA. The GSEA analysis indicated that PROM1 was negatively associated with IGF1 and mTOR pathways (Fig.13). We speculated that PROM1 might in uence metastasis BRCA development through negatively regulating IGF1 and mTOR pathways.

Discussion
It has been reported that 70% of patients with metastatic BRCA have a marked tendency to spread to the bone, resulting in signi cant skeletal complications and mortality [11]. Despite advances in diagnosis, the identi cation of patients at high risk of bone recurrence is still an unmet clinical need. Therefore, identifying useful biomarkers was conducive to improve the clinical outcome of metastatic BRCA patients. Spadazzi C et al. have identi ed TFF1 as strictly correlated to bone metastasis from ER+ breast cancer, and TFF1 upregulation could be useful to identify patients at high risk of bone metastasis [12]. Pantano F et al. determined ITGA5 as a predictive of poor bone metastasis-free survival [8]. As bone is the most frequent organ for breast cancer metastasis, and thus it is essential to predict the bone metastasis of breast cancer.
In this study, we rstly identi ed 74 signi cant DEGs between BRCA samples with or without bone metastasis. These 74 DEGs mainly participated in ECM-receptor interaction and Cytokine-cytokine receptor interaction, which referred to signaling molecules and interaction. The previous study has indicated that ECM-receptor interaction signi cantly participated in breast cancer metastasis to bone [13] and brain [14]. ECM-receptor interaction pathway was also correlated to lung metastasis in osteosarcomas [15], lung adenocarcinoma metastasis [16] and liver metastasis of colorectal cancer [17]. In addition, virus infection such as human papillomavirus (HPV) infection was also proved to correlated with DEGs in our study. During the last decades, great interest has been given to the viral etiology of breast cancer. Habyarimana et al. showed that HPV DNA was found in 46.81% of Rwandese breast cancer cases, HPV16 being the most prevalent subtype (77.27%) followed by HPV33 (13.64%) and HPV31 (9.09%) [18], suggesting high-risk HPV infections as a risk factor in breast cancer development. Cavalcante et al. found that the high frequency of HPV infection in breast cancer samples indicated a potential role in breast carcinogenesis [19]. However, Hedau et al. found no evidence of HPV etiology of breast cancer in Indian women [20]. And the low frequency of HPV was detected in Ghaffari et al. study, which also did not support the association between breast carcinoma and HPV infection [21]. It followed that effect of HPV infection in BRCA did not be uniform, and it was possible that HPV may be responsible for breast carcinogenesis only in a small percentage of all breast cancer.
Subsequently, we found 5 more important hub genes among all GEGs, namely MMP9, ITGB3, BMP2, CCL2 and PROM1. A univariate cox regression initially identi ed that PROM1 and CCL2 expressions were related to the prognosis of metastatic BRCA patients. Survival analysis, nomogram, and ROC curve analyses further presented the signi cance of PROM1 on the survival probability in metastatic BRCA.
Expression analysis showed that PROM1 was upregulated in BRCA bone metastasis samples compared with no bone metastasis samples. The above results showed that PROM1 was an important DEGs involved in bone metastasis of BRCA, and presented a good prediction performance on the patient's survival probability.
Prominin-1 (PROM1) is a transmembrane glycoprotein, which is expressed in stem cell lineages [22]. Katoh et al. determined the PROM1 as a functional cancer stem cells (CSC) marker [23]. At present, there were few kinds of research reporting the role of PROM1 in BRCA. Priedigkeit et al. reported that PROM1 was related to transcriptional remodeling in long-term estrogen-deprived locoregional breast cancer recurrences [24]. Bertheau et al. identi ed PROM1 as a mutant TP53-associated gene involved in breast cancer [25]. The important role of PROM1 in BRCA needs more investigation in the future. This study has indicated the important function of PROM1 in BRCA metastasis, and we nally explored the potential mechanism associated with PROM1. GSEA analysis indicated that PROM1 was signi cantly related to mTOR and IGF1 pathways with a negative correlation. The mTOR critically regulated several essential biological functions, such as cell growth, metabolism, survival, and immune response [26]. However, the mTOR was frequently deregulated in human cancers [27]. It has demonstrated that PROM1 played a key role in the regulation of autophagy via upstream suppression of mTOR signaling in the human retinal pigment epithelium [22]. Kholodenko et al. found that cells with the complete knockout of PROM1 showed the highest resistance to mTOR inhibitors in colorectal cancer [28]. Our study suggested that low expression of PROM1 might activate the mTOR pathway in BRCA metastasis samples. Insulin-like growth factors (IGF) are the most abundant growth factors in bone and are required for normal skeletal development and function. Via activation of the IGF-1 receptors (IGF-1R) and variant insulin receptors, IGFs promote cancer progression, aggressiveness, and treatment resistance [29]. Preclinical evidence has suggested that a high IGF-1 environment in primary tumor stimulated tumor cells metastasis to bone, suggesting that bone metastases may re ect IGF dependency [29]. Several studies have indicated that IGF signaling systems were able to regulate BRCA growth, progression, and metastasis [30,31]. It followed that the IGF1 pathway played a vital function in the progression of BRCA metastasis. However, this study just indicated the negative correlation between PROM1 and IGF1/mTOR pathways, the detailed regulation in BRCA metastasis especially bone metastasis was worth further investigating.

Conclusion
This study identi ed 74 DEGs between BRCA samples with or without bone metastasis, and 5 important DEGs were then ltered, namely MMP9, ITGB3, BMP2, CCL2 and PROM1. Through Cox regression and Kaplan-Meier survival analysis, PROM1 and CCL2 were determined as the signi cant prognosis-related biomarkers associated with metastatic BRCA. And the risk model constructed by these 2 genes showed a favorable prediction property on patient's survival. Prognosis veri cation analysis further indicated the importance of PROM1 rather than CCL2. Especially, both nomogram and ROC analyses presented the better prediction ability of PROM1 on the survival probability of metastatic BRCA patients. Finally, we found that PROM1 negatively correlated with IGF1 and mTOR pathways involved in BRCA metastasis. This study identi ed PROM1 as an important prognosis-related biomarker associated with metastatic BRCA, and detailed function needed further investigation via experimental veri cation and clinical cohort.    The PPI network among DEGs. Green nodes indicated the top 10 hub genes identi ed by the Degree method. Abbreviation: DEGs, differential expressed genes; PPI, protein-protein interaction network.

Figure 5
The mRNA expression veri cation of 5 hub genes between BRCA bone metastasis and non-bone metastasis samples in GSE14020 dataset. Abbreviation: BRCA, breast cancer.