Distinct Clinical Impact and Biological Function of Angiopoietin and Angiopoietin-like Proteins in Human Breast Cancer

Secreted angiopoietin/angiopoietin-like (ANGPT/ANGPTL) proteins are involved in many biological processes. However, the role of these proteins in human breast cancers (BCs) remains largely unclear. Here, we conducted integrated omics analyses to evaluate the clinical impact of ANGPT/ANGPTL proteins and to elucidate their biological functions. In BCs, we identified rare mutations in ANGPT/ANGPTL genes, frequent gains of ANGPT1, ANGPT4, and ANGPTL1, and frequent losses of ANGPT2, ANGPTL5, and ANGPTL7, but observed that ANGPTL1, 2, and 4 were robustly downregulated in multiple datasets. The expression levels of ANGPTL1, 5, and 8 were positively correlated with overall survival (OS), while the expression levels of ANGPTL4 were negatively correlated with OS. Additionally, the expression levels of ANGPTL1 and 7 were positively correlated with distant metastasis-free survival (DMFS), while the expression levels of ANGPT2 and ANGPTL4 were negatively correlated with DMFS. The prognostic impacts of ANGPT/ANGPTL genes depended on the molecular subtypes and on clinical factors. We discovered that various ANGPT/ANGPTL genes were co-expressed with various genes involved in different pathways. Finally, with the exception of ANGPTL3, the remaining genes showed significant correlations with cancer-associated fibroblasts, endothelial cells, and microenvironment score, whereas only ANGPTL6 was significantly correlated with immune score. Our findings provide strong evidence for the distinct clinical impact and biological function of ANGPT/ANGPTL proteins, but the question of whether some of them could be potential therapeutic targets still needs further investigation in BCs.


Introduction
Breast cancer (BC) is one of the leading causes of death among women worldwide [1][2][3]. It is well known that BC is a complex and heterogeneous disease with substantial variation in its molecular and clinical characteristics [4,5]. Multi-omics technologies have proved to be invaluable tools for deconvoluting the heterogeneity and complexity of somatic BC genetics, providing a tremendous amount of information relating to the definition of new biomarkers for diagnosis, prognosis, and the prediction of therapeutic response and to the identification of new potential therapeutic targets. Based on these findings, a few genomic prognostic tests are available for BC, such as Oncotype Dx (Genomic Health Inc., Redwood City, CA, USA) and MammaPrint (Agendia, Amsterdam, The Netherlands). However, clinical information. The GSE96058 dataset was downloaded from the Gene Expression Omnibus (GEO) database, while the METABRIC and TCGA datasets were downloaded from cBioPortal. Univariate and multivariate Cox regression analyses were executed in these three datasets using SPSS.
The tumor immune infiltration scores, stroma scores, microenvironment scores, cancerassociated fibroblasts, and endothelial cells in TCGA were downloaded from TIMER2.0 (http://timer.cistrome.org/) on 1 October 2020 [29], and were enumerated from transcriptomes using the xCell method, a novel gene-signature-based method [30]. The Spearman correlations between the expression of ANGPT/ANGPTL genes and these biological factors in the TCGA-BRCA data were calculated using SPSS (IBM SPSS statistics version 24).

Genomic Alterations in ANGPT/ANGPTL Genes in Breast Cancers
To gain insight into understanding the role of ANGPT/ANGPTL genes in human BC development and progression, we first investigated their genomic alterations in BCs. Upon mining the TCGA-BRCA data, we observed low mutational frequency without hotspots in ANGPT/ANGPTL genes. These observations were further verified by the frequencies reported in the COSMIC database ( Table 1), indicating that ANGPT/ANGPTL genes are rarely mutated in BCs. Next, we investigated the changes in the transcriptional levels of ANGPT/ANGPTL genes by comparing their expression profiles in normal breast and BC tissues using GEO2R. We observed that ANGPTL1, 2, and 4 were robustly and significantly downregulated in invasive ductal carcinoma (IDC) across all microarray datasets in the GEO database ( Figure 1A, Table S1). However, we discovered that the transcriptional levels of all ANGPT/ANGPTL genes were significantly lower in BCs than in normal breast tissues in the TCGA dataset ( Figure 1B-L). Moreover, the significant downregulation of ANGPT2, ANGPT2, ANGPTL1, ANGPTL4, and ANGPTL6 was found in tumor-adjacent tissues ( Figure 1B-L). Interestingly, the downregulation of ANGPT1 and ANGPTL1, 2, and 4 was found in ductal carcinoma in situ (DCIS) in one dataset ( Figure 1A, Table S1). To search for the possible mechanism by which the transcriptional levels of the ANGPT/ANGPTL genes were altered in BCs, we examined the DNA copy number changes of the ANGPT/ANGPTL genes in the TCGA-BRCA database and found a frequent increase in ANGPT1, ANGPT4, and ANGPTL1 and a frequent decrease in ANGPT2, ANGPTL5, and ANGPTL7 in BCs (Figure 2, left panel). Surprisingly, we discovered that the transcriptional expression levels were not significantly associated with their copy numbers, except in the case of ANGPTL3 (Figure 2, right panel). These findings indicate that DNA copy number changes do not contribute to the downregulation of ANGPT/ANGPTL genes, suggesting that their expression is mainly controlled by other mechanisms such as methylation and the regulation of transcriptional factors. ANGPT1, ANGPT4, and ANGPTL1 and a frequent decrease in ANGPT2, ANGPTL5, and ANGPTL7 in BCs (Figure 2, left panel). Surprisingly, we discovered that the transcriptional expression levels were not significantly associated with their copy numbers, except in the case of ANGPTL3 (Figure 2, right panel). These findings indicate that DNA copy number changes do not contribute to the downregulation of ANGPT/ANGPTL genes, suggesting that their expression is mainly controlled by other mechanisms such as methylation and the regulation of transcriptional factors.

Prognostic Impact of ANGPT/ANGPTL Genes in Breast Cancer Patients
To investigate whether transcriptional levels of individual ANGPT/ANGPTL genes were associated with OS, we conducted a meta-analysis using bc-GenExMiner v4.6. A meta-analysis of microarray data revealed that the expression levels of ANGPTL1, 5, and 8 positively correlated with the OS, while the expression levels of ANGPT2 and ANGPTL4 negatively correlated with the OS in BC patients (p < 0.05, Figure 3, Figure S1). Using RNA-

Prognostic Impact of ANGPT/ANGPTL Genes in Breast Cancer Patients
To investigate whether transcriptional levels of individual ANGPT/ANGPTL genes were associated with OS, we conducted a meta-analysis using bc-GenExMiner v4.6. A meta-analysis of microarray data revealed that the expression levels of ANGPTL1, 5, and 8 positively correlated with the OS, while the expression levels of ANGPT2 and ANGPTL4 negatively correlated with the OS in BC patients (p < 0.05, Figures 3 and S1). Using RNA-seq data, we found that the expression levels of ANGPT4 and ANGPTL1, 5, 7, and 8 positively correlated with the OS, while the expression levels of ANGPTL4 negatively correlated with the OS in BC patients (p < 0.05, Figures 3 and S1). These findings indicated that only ANGPTL1, 4, 5, and 8 are consistently associated with OS in both the microarray and RNA-seq data. seq data, we found that the expression levels of ANGPT4 and ANGPTL1, 5, 7, and 8 positively correlated with the OS, while the expression levels of ANGPTL4 negatively correlated with the OS in BC patients (p < 0.05, Figure 3, Figure S1). These findings indicated that only ANGPTL1, 4, 5, and 8 are consistently associated with OS in both the microarray and RNA-seq data.  Some studies report that some of the ANGPT/ANGPTL genes play a critical role in tumor progression [12,15]. Next, we assessed the association between ANGPT/ANGPTL genes and DMFS using bc-GenExMiner, where some microarray studies contain the DMFS information. A meta-analysis showed that ANGPTL1 and 7 were positively correlated with the DMFS, while the expression levels of ANGPT2 and ANGPTL4 were negatively correlated with the DMFS (Figures 4 and S2). All these findings support the evidence that some ANGPT/ANGPTL genes have a prognostic impact in BCs. Some studies report that some of the ANGPT/ANGPTL genes play a critical role in tumor progression [12,15]. Next, we assessed the association between ANGPT/ANGPTL genes and DMFS using bc-GenExMiner, where some microarray studies contain the DMFS information. A meta-analysis showed that ANGPTL1 and 7 were positively correlated with the DMFS, while the expression levels of ANGPT2 and ANGPTL4 were negatively correlated with the DMFS (Figure 4, Figure S2). All these findings support the evidence that some ANGPT/ANGPTL genes have a prognostic impact in BCs.

Molecular-Subtype-Dependent Prognostic Impact of ANGPT/ANGPTL Genes in Breast Cancers
The molecular subtype is an important prognostic factor in BCs. Therefore, we examined whether stratifying tumors according to their molecular subtype could reveal additional information about the association between ANGPT/ANGPTL genes and BCs. First, each patient was assigned to a molecular subtype based on PAM50 [31]. The frequencies of the copy number changes in ANGPT/ANGPTL genes were found to be significantly different in different molecular subtypes ( Figure S3). We then performed an impact analysis of ANGPT/ANGPTL genes on the OS and DMFS of patients in each molecular subtype and found that the association between ANGPT/ANGPTL genes and the OS and DMFS strongly depended on the molecular subtype ( Figure 5). For example, significant association between transcriptional levels of ANGPTL1 and the OS and DMFS was only found in the basal type ( Figure 5).

Molecular-Subtype-Dependent Prognostic Impact of ANGPT/ANGPTL Genes in Breast Cancers
The molecular subtype is an important prognostic factor in BCs. Therefore, we examined whether stratifying tumors according to their molecular subtype could reveal additional information about the association between ANGPT/ANGPTL genes and BCs. First, each patient was assigned to a molecular subtype based on PAM50 [31]. The frequencies of the copy number changes in ANGPT/ANGPTL genes were found to be significantly different in different molecular subtypes ( Figure S3). We then performed an impact analysis of ANGPT/ANGPTL genes on the OS and DMFS of patients in each molecular subtype and found that the association between ANGPT/ANGPTL genes and the OS and DMFS strongly depended on the molecular subtype ( Figure 5). For example, significant association between transcriptional levels of ANGPTL1 and the OS and DMFS was only found in the basal type ( Figure 5). To assess the prognostic impact of ANGPT/ANGPTL genes independently of clinical factors and molecular subtypes, we checked all available datasets and only found three datasets that contained both the transcriptional data of all ANGPT/ANGPTL genes and data on clinical factors. Consistently with the findings of the meta-analysis described above, some ANGPT/ANGPTL genes showed significant association with the OS according to univariate Cox regression ( Figure 6). However, multivariate Cox regression analyses (including age, pathological stage, ER status, PR status, tumor size, and molecular subtype) were only significant in one dataset after adjusting for clinical factors and molecular subtypes ( Figure 6).
Taken together, our findings suggest that the prognostic impacts of ANGPT/ANGPTL genes are remarkably dependent on clinical factors and molecular subtypes. To assess the prognostic impact of ANGPT/ANGPTL genes independently of clinical factors and molecular subtypes, we checked all available datasets and only found three datasets that contained both the transcriptional data of all ANGPT/ANGPTL genes and data on clinical factors. Consistently with the findings of the meta-analysis described above, some ANGPT/ANGPTL genes showed significant association with the OS according to univariate Cox regression ( Figure 6). However, multivariate Cox regression analyses (including age, pathological stage, ER status, PR status, tumor size, and molecular subtype) were only significant in one dataset after adjusting for clinical factors and molecular subtypes ( Figure 6).
Taken together, our findings suggest that the prognostic impacts of ANGPT/ANGPTL genes are remarkably dependent on clinical factors and molecular subtypes.

Biological Functions of ANGPT/ANGPTL Genes in Breast Cancers Elucidated via Gene Co-Expression Network
Although many studies have revealed various functions of ANGPT/ANGPTL genes [14,15], to obtain further insight into their differences with respect to the underlying mechanisms of tumor development and progression a co-expression analysis of individual ANGPT/ANGPTL genes was performed for the RNA-seq data using bc-GenExMiner. A number of genes that are significantly co-expressed with ANGPT/ANGPTL genes are shown in Table 2 (|r| ≥ 0.40; p < 1.00 × 10 −4 ). Distinct sets of the genes were co-expressed with ANGPT/ANGPTL genes (Figure 7, Table S2). Gene Ontology (GO) functional enrichment analysis of these co-expressed genes showed significant enrichment for the distinct biological processes involved for individual ANGPT/ANGPTL genes (adjusted p-value < 0.05, Figure 8A, Figure S4, Table S3). Not surprisingly, it was found that the genes that were positively correlated with ANGPT1, 2, and 4 and ANGPTL1 and 5 were significantly enriched for the biological processes involved in angiogenesis ( Figure 8A). This analysis also revealed that ANGPT1, 2, and 4 and ANGPTL1 and 2 are possibly involved in regulating the extracellular matrix (ECM), ANGPTL6 possibly has a function in the regulation of immunity, and ANGPTL4 and 8 possibly regulate lipid metabolism ( Figure 8A). Additionally, those genes negatively correlated with ANGPT4 and ANGPTL1 were significantly enriched for biological processes involved in the cell cycle ( Figure S3). Moreover, KEGG analysis indicated that the co-expressed genes were significantly enriched for the distinct pathways involved by ANGPT/ANGPTL genes ( Figure 8B, Figure S5, Table S4). These findings indicate distinct molecular mechanisms associated with ANGPT/ANGPTL genes in breast tumor development and progression.

Correlation of ANGPT/ANGPTL Genes with Biological Factors in the Tumor Microenvironment of Breast Cancers
The tumor microenvironment, which contains infiltrating host cells, secreted factors, and extracellular matrix proteins, profoundly influences tumor progression and therapeutic responses [32]. Therefore, finally, we assessed the correlations between ANGPT/ANGPTL genes and biological factors in the tumor microenvironment of breast cancers using TCGA data (Table S5). Consistently with the biological function enrichment analysis of the co-expressed genes, ANGPTL6 was strongly and significantly correlated

Biological Functions of ANGPT/ANGPTL Genes in Breast Cancers Elucidated via Gene Co-Expression Network
Although many studies have revealed various functions of ANGPT/ANGPTL genes [14,15], to obtain further insight into their differences with respect to the underlying mechanisms of tumor development and progression a co-expression analysis of individual ANGPT/ANGPTL genes was performed for the RNA-seq data using bc-GenExMiner. A number of genes that are significantly co-expressed with ANGPT/ANGPTL genes are shown in Table 2 (|r| ≥ 0.40; p < 1.00 × 10 −4 ). Distinct sets of the genes were co-expressed with ANGPT/ANGPTL genes (Figure 7, Table S2). Gene Ontology (GO) functional enrichment analysis of these co-expressed genes showed significant enrichment for the distinct biological processes involved for individual ANGPT/ANGPTL genes (adjusted p-value < 0.05, Figures 8A and S4, Table S3). Not surprisingly, it was found that the genes that were positively correlated with ANGPT1, 2, and 4 and ANGPTL1 and 5 were significantly enriched for the biological processes involved in angiogenesis ( Figure 8A). This analysis also revealed that ANGPT1, 2, and 4 and ANGPTL1 and 2 are possibly involved in regulating the extracellular matrix (ECM), ANGPTL6 possibly has a function in the regulation of immunity, and ANGPTL4 and 8 possibly regulate lipid metabolism ( Figure 8A). Additionally, those genes negatively correlated with ANGPT4 and ANGPTL1 were significantly enriched for biological processes involved in the cell cycle ( Figure S3). Moreover, KEGG analysis indicated that the co-expressed genes were significantly enriched for the distinct pathways involved by ANGPT/ANGPTL genes ( Figures 8B and S5, Table S4). These findings indicate distinct molecular mechanisms associated with ANGPT/ANGPTL genes in breast tumor development and progression.     Table S2.

Correlation of ANGPT/ANGPTL Genes with Biological Factors in the Tumor Microenvironment of Breast Cancers
The tumor microenvironment, which contains infiltrating host cells, secreted factors, and extracellular matrix proteins, profoundly influences tumor progression and therapeutic responses [32]. Therefore, finally, we assessed the correlations between ANGPT/ANGPTL genes and biological factors in the tumor microenvironment of breast cancers using TCGA data (Table S5). Consistently with the biological function enrichment analysis of the coexpressed genes, ANGPTL6 was strongly and significantly correlated with the immune score (Table 3). Except for ANGPTL3, the remaining ANGPT/ANGPTL genes were significantly correlated with the stroma and microenvironment scores, cancer-associated fibroblasts, and endothelial cells (Table 3). These findings suggest that the contribution of ANGPT/ANGPTL genes to BC development and progression may be through the regulation of microenvironments. Cells 2021, 10, x FOR PEER REVIEW 11 of 15

Discussion
It is well known that tumor metastasis is the real culprit and underlying cause of most BC-related deaths [1]. It is urgently necessary to design and develop effective therapeutics to block metastases. In this study, we used multiple bioinformatics tools to delineate the potential roles of 11 ANGPT/ANGPTL genes in BC since few of them have been well studied. However, ANGPT2 has been shown to play an important role in BC in many studies [33][34][35][36][37]. The ANGPT/ANGPTL proteins play a critical role in the regulation of cancer angiogenesis, which is an essential process for tumor metastasis [6,8,9]. Summarizing our findings, we conclude that ANGTPL1 and 4 are the most promising potential targets with respect to BC, although further investigations are still needed, as we discuss in detail below.
We robustly observed that ANGTPL1 and 4 were significantly downregulated in BCs, and their expression levels were significantly associated with the OS and DMFS of patients. In contrast, for the others, significance was only found in a subset of the data. One study showed that ANGPTL1 inhibits BC cell migration and invasion in vitro [16]. It is worth noting that transcriptome profiling of metastatic canine mammary carcinomas shows the significant downregulation of ANGPT2 and ANGPTL1-4 compared to normal mammary glands [38]. Consistently with these results, our study shows that high expression levels of ANGPTL1 significantly prolong the DMFS of BC patients. The co-expression network and function enrichment analysis revealed that in addition to the regulation of angiogenesis as a key essential anti-angiogenic protein [13], ANGPTL1 affects ECM regulation and suppresses cell cycles. These results suggest that ANPTL1 plays a tumor-suppressive role in BC. Studies of other cancer types support these results. It has been reported that the ANGPTL1 transcript is downregulated in lung, prostate, kidney, thyroid, and urinary bladder cancer [39], and that ANGPTL1 suppresses metastasis in hepatocellular carcinoma [17,18], colorectal cancer [19][20][21], and lung cancer [16]. Therefore, ANGPTL1 acts as a general tumor suppressor gene in human cancers.
In addition to angiogenesis, ANGPTL4 has been reported to be involved in the regulation of lipoprotein metabolism [40]. We demonstrated that ANGPTL4 is co-expressed with well-known genes involved in lipid metabolism. Moreover, many studies have reported the involvement of ANGPTL4 in BCs. ANGPTL4 is transcriptionally regulated by TGFβ and serves as an important mediator for TGFβ1 to prime BCs for lung metastasis [41] and TGFβ2-induced BC brain metastasis [42]. The depletion of ANGPTL4 inhibits obesityinduced angiogenesis and tumor growth [43]. Consistently with these reports, we found that a high level of ANGPTL4 significantly shortens the DMFS of BC patients. One study showed that ANPTL4 is an independent poor prognostic factor for the OS and diseasefree survival (DFS) of BC patients [44]. We also observed that high levels of ANGPTL4 significantly shorten the OS of BC patients. It is worth noting that there are contradictory data in the literature about its expression alteration and its functions in human cancers. For example, a recent study demonstrated that ANGPTL4 inhibits cell migration and that high levels of ANGPTL4 prolong the OS and DFS of patients with triple-negative BC [45].
However, many studies of other types of cancer suggest that ANGPTL4 functions as an oncogene [46][47][48]. It is possible that these discrepancies are due to alternative splicing of ANGPTL4. In addition, these contradictory findings suggest a multifaceted role for ANGPTL4 in human cancers. Therefore, further investigation is required into ANGPTL4 regulatory circuits and the definition of specific molecular events that mediate its various biological functions in different cancer stages.
A limitation of this study is that all conclusions were based on bioinformatics analyses, which require to be verified by experimental and clinical studies. Nevertheless, our study uncovered the importance of ANGPT/ANGPTL genes in BC development and progression and can guide future research.

Conclusions
Our findings provide strong evidence for the distinct clinical impacts and biological functions of ANGPT/ANGPTL proteins in BC development and progression, suggesting that some of them, such as ANGPTL1 and 4, could be potential therapeutic targets for BCs.

Supplementary Materials:
The following are available online at https://www.mdpi.com/article/10 .3390/cells10102590/s1, Figure S1: Association of ANGPT1, 2 and 4, and ANGPTL2, 3, 6, and 7 with OS in both microarray and RNA-seq data. Figure S2: Association of ANGPT1 and 4, and ANGPTL2, 3, 6, and 8 with DMFS in microarray data. Figure S3: Heatmap presentation of top ten biological processes identified by GO functional enrichment analysis of the genes negatively co-expressed with angiopoietin/angiopoietin-like genes. Figure S4: Heatmap presentation of top ten pathways identified by KEGG analysis of the genes negatively co-expressed with angiopoietin/angiopoietin-like genes. Figure S5: Heatmap presentation of top ten pathways identified by KEGG analysis of the genes negatively co-expressed with ANGPT / ANGPTL genes. Table S1: Fold change (FC) and adjusted p-value (adj.P.Val)for angiopoietin/angiopoietin-like genes in different GEO datasets generated by GEO2R. Table S2: List of the genes significantly co-expressed with angiopoietin/angiopoietin-like genes in RNA-seq data. Table S3: Gene ontology (GO) functional enrichment analysis of the genes significantly co-expressed with angiopoietin/angiopoietin-like genes in RNA-seq data for biological processes. *adj.P.Val = adjusted P-Value. Table S4: Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis of the genes significantly co-expressed with angiopoietin/angiopoietinlike genes in RNA-seq data. *adj.P.Val = adjusted P-Value. Table S5 Funding: This research and the APC were funded by DOD BCRP (grant number BC190820).

Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable. Data Availability Statement: All data used in the study were downloaded from publicly available databases.