Identification of differentially expressed MAGs in BC
The mRNA expression data of TCGA BC patients were subjected to identify differentially expressed MAGs. We identified 168 differentially expressed MAGs (68 downregulated and 100 upregulated) between 414 BC tissues and 19 normal tissues. To investigate the differences in the protein levels of these 168 MAGs between BC and normal samples, we collected samples from 10 patients in the Affiliated Hospital of Qingdao University. These samples were processed and analyzed using LC-MS/MS following the process outlined in Additional file 1. The result showed that the protein levels of 23 MAGs were differentially expressed in BC among these 168 MAGs. Boxplots were used to screen the mRNA and protein levels 23 MAGs in BC (Fig. 1a, b). The detailed information of the protein expression of 23 MAGs in our samples was shown in Additional file 2. In addition, a PPI network of 23 MAGs was retrieved from the STRING database and their correlations were further screened in Cytoscape. As a result, we found AKR1B1 was a hub gene in the interaction network by MCODE analysis in Cytoscape (Fig. 1c). To identify the potential mechanisms of these 23 differentially expressed MAGs in BC, we performed GO and KEGG analyses. We found that the most significant GO enriched terms involved in metabolism were amino acid metabolic process, antibiotic metabolic process, organic hydroxy compound catabolic process, alcohol metabolic process, and glycogen catabolic process (BP, biological process); myelin sheath and mitochondrial matrix (CC, cellular component); and lyase activity, oxidoreductase activity cofactor binding, coenzyme binding electron transfer activity, and NAD binding (MF, molecular function) (Fig. 2a). In the KEGG enrichment analysis, these MAGs were primarily correlated with pathways related to arginine and proline metabolism, lysine degradation, glycerolipid metabolism, histidine metabolism, pyruvate metabolism, tryptophan metabolism, ether lipid metabolism, and glycolysis (Fig. 2b).
Identification of prognosis-related MAGs and construction of a MAGs based prognostic signature
By performing univariate Cox regression analysis on 23 MAGs, a total of 5 MAGs were identified to have significant prognostic value in BC (P<0.05) (Fig. 3a). Five MAGs were considered as risk factors with HR values greater than 1. Subsequently, we utilized multivariate Cox regression analysis to construct a prognostic signature, which contained five MAGs, including PLOD1 (procollagen-lysine,2-oxoglutarate 5-dioxygenase 1), CKB (creatine kinase B), PYGB (glycogen phosphorylase B), AKR1B1 (aldo-keto reductase family 1 member B), and PDE5A (phosphodiesterase 5A) (Table 2). We further calculated the prognostic risk score for each BC patient as follows: risk score = (0.0049 × expression level of PLOD1) + (0.0018 × expression level of CKB) + (0.0033× expression level of PYGB) + (0.0031× expression level of AKR1B1) + (0.0486 × expression level of PDE5A). Four hundred and three BC patients were subdivided into high-risk and low-risk groups according to the median value of risk score. K-M survival curve analysis showed that the MAGs based signature was closely associated with poor overall survival (OS) (P = 5.562e-05), disease-specific survival (DSS) (P = 4.896e-03), and progression-free interval (PFI) (P = 2.915e-02) in BC (Fig. 3b-d). However, the signature risk score was not correlated with the disease-free interval (DFI) (P = 7.724e-01) of BC patients (Fig. 3e). ROC curve analysis was used to further measure the predictive performance of the MAGs based signature risk score. The area under the curves (AUCs) for the MAGs signature, age, gender, grade, stage, T, M, N were 0.766, 0.549, 0.436, 0.553, 0.648, 0.623, 0.522, and 0.638, which indicated superior predictive accuracy of the MAGs signature risk score in survival outcomes (Fig. 3f). We further used univariate and multivariate Cox regression analyses to assess the prognostic values of the MAGs based signature and clinical features. Univariate Cox regression analysis showed that age, stage, T (tumor), N (node), and risk score were related to the survival of BC patients (Fig. 3g). Subsequently, multivariate Cox regression analysis indicated that the MAGs based signature was an independent prognostic factor for BC (P < 0.001, Fig. 3h).
Association between the MAGs signature risk score and clinicopathologic characteristics
The treatment methods for BC patients depend largely on clinical characteristics, and we evaluated whether there was a statistically significant difference between risk score and clinicopathological characteristics. Our study revealed that the MAGs based signature risk score was correlated with the stage (P = 1.346e-05), grade (P = 1.944e-05), T (P = 0.004), M (metastasis) (P = 0.005) of BC patients (Fig. 4a-d). However, the risk score was not related to gender (P = 0.599) and N (P= 0.086) of BC patients (Fig. 4e, f). In recent years, several independent studies have shown that BC has distinct molecular subtypes, which were associated with different outcomes of BC patients [22-24]. Therefore, the expression levels of five MAGs in two molecular types were further analyzed. The results indicated that AKR1B1, PLOD1, and PYGB were highly expressed in the basal subtype and CKB was highly expressed in the luminal subtype (Fig. 5a-d). However, the expression of PDE5A was not associated with the molecular subtypes of BC (Fig. 5e). Furthermore, MIBC patients with high MAGs based signature risk score may indicate the features of a basal subtype (Fig. 5f).
Construction of a prognostic nomogram for BC
To establish a clinically applicable method for monitoring the prognosis of BC patients, we generated a nomogram to predict the survival of BC patients, by combining age, gender, grade, stage, T, N, M with the MAGs based signature risk score. The result showed that the prognostic nomogram could superiorly predict the 1-, 2-, and 3-year survival outcomes of BC patients (Fig. 6).
Correlation analysis between the risk score and immune cell infiltration in BC
To identify the significance of the MAGs based signature in the tumor microenvironment, the relationship between the abundance of six types of tumor‐infiltrating immune cells (B cells, CD4+ T cells, CD8+ T cells, neutrophils, macrophages, and dendritic cells) and the MAGs based signature risk score was explored in BC. The results indicated that the risk score was positively associated with the infiltration of macrophages (P = 1.354e-08) and dendritic cells (P = 6.016e-04,). However, the risk score was not correlated with the infiltration of B cells, CD4+ T cells, CD8+ T cells, and neutrophils (Table 4).
M2 TAMs may promote the expression levels of MAGs via the TGF-β1 signaling pathway
RNA isolation and reverse transcription‑quantitative PCR (RT-qPCR) was further performed to validate the expression levels of five selected MAGs in T24 and SV-HUC-1 cell lines. The results demonstrated significant differences in the expression levels of five MAGs between T24 and SV-HUC-1 cell lines (Fig. 7a). Among these five MAGs, PLOD1, CKB, PYGB were upregulated, PDE5A and AKR1B1 were downregulated in T24 cells. Compared with the unstimulated T24 cells, the expression levels of five MAGs were significantly elevated in T24 cell lines after stimulated with the supernatant of M2 TAMs (Fig. 7b). Furthermore, we used the TGF-β1 inhibitor to inhibit the production of TGF-β1 in M2 TAM cells. Compare to M2 TAMs without TGF-β1 inhibitor, the production of TGF-β1 in M2 TAMs with TGF-β1 inhibitor was significantly decreased (Additional file 3). The expression levels PLOD1, CKB, and PYGB were significantly downregulated in T24 cells when stimulated with the low TGF-β1 supernatant of M2 TAMs (Fig. 7c). To investigate whether TGF-β1 alone can affect the expression of MAGs in T24 cells, the T24 cells were stimulated with recombinant human TGF-beta 1. Compared with the unstimulated T24 cells, the expression of PLOD1, CKB, and PYGB was elevated, but the expression of PDE5A and AKR1B1 was not significantly changed (Fig. 7d).
In summary, five MAGs were differentially expressed and M2 TAMs secreted TGF-β1 can promote three MAGs expression in BC cells.