3.1 Expression of eIF4E in different tumors and normal tissues
We first analyzed the expression of eIF4E in a variety of tumors and normal tissues using Oncomine database, and found that the expression of eIF4E in brain cancer, breast cancer, cervical cancer, colorectal cancer, gastric cancer, head and neck tumor, kidney cancer, lung cancer, lymphoma, ovarian cancer, pancreatic cancer, sarcoma and other tumors was higher than that in normal tissues (P <0.001) (Figure 1A). We further used the TIMER database to evaluate the differential expression of eIF4E in specific tumor types (Figure 1B). The results showed that the expression of eIF4E in invasive breast carcinoma, endometrial carcinoma, cholangiocarcinoma, colonic adenocarcinoma, hepatocellular carcinoma, gastric adenocarcinoma, lung squamous cell carcinoma, lung adenocarcinoma and esophageal carcinoma was significantly higher than that in normal controls. The expression of eIF4E in thyroid carcinoma, renal papillary cell carcinoma and renal clear cell carcinoma was significantly lower than that in normal control groups. The expression in metastatic skin melanoma was higher than that in skin melanoma (P <0.05).
Further subgroup analysis of multiple clinic-pathological features of TCGA-Breast invasive carcinoma samples in the UALCAN database consistently showed an increase in the transcriptional level of eIF4E. According to the analysis of sample type, age, subtype of breast cancer, disease stage, lymph node metastasis and TP53 mutation, the expression of eIF4E in breast cancer patients was significantly higher than that in normal controls, and the expression of eIF4E in patients aged 61 to 80 was significantly higher than that in patients aged 41 to 60, with statistical difference (P =0.037399). In all subtypes of breast cancer, the expression of eIF4E was significantly higher than that of normal subjects, and the expression of Luminal type was significantly higher than that of Triple negative type (P <0.01), and the expression levels of stage1, stage2 and stage3 in different tumor stages were significantly higher than those in normal group. Lymph node metastasis showed that the expression level of N2 was the highest and significantly different from that of N0 (P =0.0127978) and N3 (P =0.0169045). TP53 mutation analysis showed that the expression level of TP53 non-mutated group was higher than that of the mutant group (P =0.024296) (Figure 2). Therefore, according to the expression differences of breast cancer subtypes, tumor stages and lymph node metastasis, the expression of eIF4E can be used as a potential diagnostic index in BRCA.
3.2 Relationship between eIF4E expression and prognosis of patients with different tumors
Next, we used the PrognoScan database to explore the relationship between the expression of eIF4E and the prognosis of tumor patients. We found that breast and colorectal cancers were significantly associated with the expression of eIF4E (Figure 3A-B) (DSS: Disease Specific Survival; RFS: Relapse Free Survival).In addition, we used the Kaplan-Meier database to evaluate the relationship between the expression of eIF4E in a range of tumor types and prognosis. The results showed that the increased expression of eIF4E was significantly correlated with the poor prognosis of breast cancer (OS HR = 1.32, 95% CI =1.02-1.71, P =0.037; HR =1.41, 95% CI =1.27-1.857, P =5.3e-10). The increased expression of eIF4E was also significantly associated with poor prognosis in ovarian cancer (OS HR=1.1810 95%CI=1.02-1.36, P =0.026). However, in lung and gastric cancers, decreased expression of eIF4E was significantly associated with poor prognosis (lung cancer OS HR =0.86, 95%CI =0.76-0.98 P =0.019; gastric cancer OS HR =0.54, 95%CI =0.44-0.65 P =1.1e-10) (Figure 3C-G). We further used GEPIA database to evaluate the relationship between the expression of eIF4E and the prognosis of patients, and analyzed 33 tumor types. It was found that the prognosis of high expression of eIF4E was poor in breast cancer, brain low-grade glioma, lung adenocarcinoma, cervical squamous cell carcinoma and adenocarcinoma, hepatocellular carcinoma and lung squamous cell carcinoma, while the low expression of eIF4E in renal clear cell carcinoma, hepatic clear cell carcinoma and colorectal cancer had poor prognosis(Supplement figure1A-I).These results clearly showed that in many tumor types, the expression of eIF4E was significantly correlated with poor prognosis, and the high expression of eIF4E in different databases was significantly correlated with poor prognosis of breast cancer patients.
3.3 Relationship between eIF4E expression and infiltration of immune cells in breast cancer
Gene expression data set GSE109169, related to breast cancer was searched from the comprehensive gene expression database (GEO) to analyze the difference of gene expression between breast cancer and adjacent normal tissues (Supplementary table1).The abundance of immune cell infiltration was calculated by ImmuCellAI database. It was found that in 18 kinds of T cells and other 6 types of immune cells, the infiltration levels of macrophages, nTreg cells, Th1, B cells, CD8+T cells and γδT cells in breast cancer tissues were significantly higher than those in adjacent normal tissues. Infiltration levels of Th17 cells, Tfh, NKT cells, monocytes, neutrophils and CD4+T cells in tumor tissues were lower than those in normal tissues (Figure 4). This showed that there are a significant difference in immune cell infiltration between breast cancer patients and adjacent normal tissues, and different levels of immune cell infiltration have potential effects on the occurrence, development and survival of breast cancer patients.
Since we found that the expression of eIF4E was related to the poor prognosis of patients with breast cancer, we further drew a Kaplan-Meier map using the TIMER database to explore the relationship between immune cell infiltration and the expression of eIF4E to explore its potential mechanism in breast cancer. In breast cancer, eIF4E expression was significantly correlated with tumor purity (r =0.134 P =2.21e-05), CD8+T cells(r =0.268 P =1.51e-17), macrophages (r =0.237 P =5.07e-14), neutrophils(r =0.161 P =6.14e-07) and dendritic cells (r =0.067 P =3.94e-02) (Figure 5A).
In order to further study the relationship between immune cell infiltration and eIF4E expression in BRCA, we further used TIMER database to generate Kaplan-Meier map. We found that the infiltration of CD8+T cells (P =0.006), CD4+T cells (P =0.006), neutrophils (P =0.007) and dendritic cells (P =0.004) was significantly correlated with the prognosis of BRCA (Figure 5B). In addition, the multivariate hazard model was used to evaluate the effect of eIF4E expression in the presence of different immune cell infiltration. The OS risk of eIF4E was 1.482 times higher (P <0.05) (Figure 5C). This suggested that eIF4E played an important role in regulating immune cell infiltration in breast cancer.
3.4 Evaluation of the correlation between eIF4E and the expression of immune markers
Next, we used TIMER databases to further explore the relationship between the expression of eIF4E and the level of immune cell infiltration in breast cancer. We evaluated the correlation between eIF4E expression and specific cell subsets, including CD8+T cells, B cells, monocytes, TAMs, M1 and M2 macrophages, neutrophils, NK cells, DC, Th1 cells, Th2 cells, Tfh cells, Th17 cells, Treg cells and Exhaustion T cells. We adjusted these results according to the purity of the tumor. CD8+T cells (CD8B), B cells (CD19, CD79A), monocyte markers (CD86), TAM markers (CCL2, CD68, IL10), M1 macrophage markers (COX2), M2 macrophage markers (CD163, VSIG4, MS4A4A), Neutrophils markers (CD11b), DC markers (HLA-DPB1, HLA-DRA, HLA-DPA1, BCDA-4), Th1 markers (STAT4, STAT1), Th2 markers (GATA3, STAT6), Tfh markers (BCL6, IL21), Th17 (STAT3), Treg markers (FOXP3, CCR8, STAT5B), Exhaustion T cell makers(PD-1, LAG3, TIM-3) were significantly correlated with eIF4E expression (Table 1).The expression of eIF4E in breast cancer was positively correlated with the expression of monocytes, TAM, M1 macrophages, M2 macrophages, Neutrophils, Th1, Th2, Tfh, Th17 and Treg markers, and negatively correlated with CD8+T cells, B cells, dendritic cells and Exhaustion T cell markers. Figure 6A showed the scatter diagram of TAM, M2 macrophages, Th1, Th2, Th17, Treg and Exhaustion T cell markers.
Table 1
Correlation analysis between eIF4E and relate genes and markers of immune cells in TIMER.
Description
|
Gene makers
|
Breast cancer
|
None
|
Purity
|
Cor
|
P
|
|
partial.cor
|
partial.p
|
|
CD8+T cell
|
CD8A
|
-0.022
|
4.75×10-1
|
|
0.04686004
|
0.13985143
|
|
CD8B
|
-0.148
|
8.28×10-7
|
***
|
-0.10188943
|
0.00129684
|
*
|
T cell(general)
|
CD3D
|
-0.125
|
3.09×10-5
|
***
|
-0.06614794
|
0.03705509
|
*
|
CD3E
|
-0.094
|
1.82×10-3
|
*
|
-0.02731704
|
0.38961155
|
|
CD2
|
-0.043
|
1.5×10-1
|
|
0.02692737
|
0.3964128
|
|
B cell
|
CD19
|
-0.146
|
1.08×10-6
|
***
|
-0.09601416
|
0.00244306
|
*
|
CD79A
|
-0.136
|
5.99×10-6
|
***
|
-0.0761307
|
0.01636396
|
*
|
Monocyte
|
CD86
|
0.096
|
1.47×10-3
|
*
|
0.15738691
|
6.1348E-07
|
***
|
CD115(CSF1R)
|
0.001
|
9.73×10-1
|
|
0.0607296
|
0.05561681
|
|
TAM
|
CCL2
|
0.002
|
9.39×10-1
|
|
0.06986978
|
0.02761177
|
*
|
CD68
|
0.065
|
3.01×10-2
|
|
0.11656366
|
0.00023057
|
**
|
IL10
|
0.153
|
3.38×10-7
|
***
|
0.21721067
|
4.4418E-12
|
***
|
M1 Macrophage
|
INOS(NOS2)
|
-0.024
|
4.34×10-1
|
|
-0.0130669
|
0.68072849
|
|
IRF5
|
0.016
|
5.91×10-1
|
|
0.03883945
|
0.22116287
|
|
COX2(PTGS2)
|
0.043
|
1.56×10-1
|
|
0.12822816
|
5.0261E-05
|
***
|
M2 Macrophage
|
CD163
|
0.152
|
4.1×10-7
|
***
|
0.20522705
|
6.4959E-11
|
***
|
VSIG4
|
0.077
|
1.04×10-2
|
|
0.12544706
|
7.3161E-05
|
***
|
MS4A4A
|
0.138
|
4.39×10-6
|
***
|
0.210251
|
2.1511E-11
|
***
|
Neutrophils
|
CD66b(CEACAM8)
|
-0.018
|
5.52×10-1
|
|
-0.01102645
|
0.72843176
|
|
CD11b(ITGAM)
|
0.041
|
1.76×10-1
|
|
0.08676308
|
0.00619708
|
*
|
CCR7
|
-0.055
|
6.61×10-2
|
|
0.0154633
|
0.62630164
|
|
Natural killer cell
|
KIR2DL1
|
-0.015
|
6.12×10-1
|
|
0.00357961
|
0.9102562
|
|
KIR2DL3
|
-0.008
|
7.97×10-1
|
|
0.0146039
|
0.64560698
|
|
KIR2DL4
|
-0.027
|
3.62×10-1
|
|
0.00685601
|
0.8290771
|
|
KIR3DL1
|
-0.04
|
1.82×10-1
|
|
-0.00907876
|
0.77497103
|
|
KIR3DL2
|
-0.051
|
9.37×10-2
|
|
-0.01393341
|
0.66083692
|
|
KIR3DL3
|
0.004
|
8.82×10-1
|
|
0.01473616
|
0.64262004
|
|
KIR2DS4
|
-0.028
|
3.57×10-1
|
|
0.01435843
|
0.65116604
|
|
Dendritic cell
|
HLA-DPB1
|
-0.167
|
2.46×10-8
|
***
|
-0.11866431
|
0.000177
|
**
|
HLA-DQB1
|
-0.113
|
1.75×10-4
|
**
|
-0.05129112
|
0.10606764
|
|
HLA-DRA
|
0.011
|
7.1×10-1
|
|
0.08384234
|
0.0081765
|
*
|
HLA-DPA1
|
0.011
|
7.12×10-1
|
|
0.08372552
|
0.00826629
|
*
|
BCDA-1(CD1C)
|
-0.083
|
5.95×10-3
|
*
|
-0.0070813
|
0.82355157
|
|
BCDA-4(NRP1)
|
0.144
|
1.59×10-6
|
***
|
0.20736387
|
4.0736E-11
|
***
|
CD11c(ITGAX)
|
-0.02
|
5.17×10-1
|
|
0.04743547
|
0.1350466
|
|
Th1
|
T-bet(TBX21)
|
-0.096
|
1.45×10-3
|
*
|
-0.03946619
|
0.21379428
|
|
STAT4
|
0.024
|
4.34×10-1
|
|
0.10698215
|
0.00072919
|
**
|
STAT1
|
0.275
|
1.62×10-20
|
***
|
0.29435209
|
2.558E-21
|
***
|
IFN-γ(IFNG)
|
-0.033
|
2.75×10-1
|
|
0.0081502
|
0.79745787
|
|
TNF-α(TNF)
|
-0.068
|
2.51×10-2
|
|
-0.03221895
|
0.31021222
|
|
Th2
|
GATA3
|
0.301
|
1.66×10-24
|
***
|
0.27086187
|
3.5639E-18
|
***
|
STAT6
|
0.075
|
1.32×10-2
|
|
0.10105935
|
0.00142107
|
*
|
STAT5A
|
0.001
|
9.78×10-1
|
|
0.04334184
|
0.17212936
|
|
IL13
|
-0.009
|
7.67×10-1
|
|
0.02091229
|
0.51017828
|
|
Tfh
|
BCL6
|
0.062
|
4.01×10-2
|
|
0.10660096
|
0.00076196
|
**
|
IL21
|
0.041
|
1.71×10-1
|
|
0.07605566
|
0.01647037
|
*
|
Th17
|
STAT3
|
0.304
|
5.92×10-25
|
***
|
0.31529031
|
2.2229E-24
|
***
|
IL17A
|
-0.007
|
8.23×10-1
|
|
0.01164907
|
0.71375374
|
|
Treg
|
FOXP3
|
0.006
|
8.53×10-1
|
|
0.07298812
|
0.02137316
|
*
|
CCR8
|
0.217
|
3.19×10-13
|
***
|
0.27040878
|
4.0704E-18
|
***
|
STAT5B
|
0.254
|
1.35×10-17
|
***
|
0.27303179
|
1.8794E-18
|
***
|
TGFβ(TGFB1)
|
-0.093
|
1.95×10-3
|
*
|
-0.0489941
|
0.12267125
|
|
cell exhaustion
|
PD-1(PDCD1)
|
-0.142
|
2.14×10-6
|
***
|
-0.09726802
|
0.00214011
|
*
|
CTLA4
|
-0.037
|
2.15×10-1
|
|
0.02018994
|
0.52490196
|
|
LAG3
|
-0.14
|
3.09×10-6
|
***
|
-0.11585047
|
0.00025198
|
**
|
TIM-3(HAVCR2)
|
0.133
|
9.24×10-6
|
***
|
0.18444976
|
4.6761E-09
|
***
|
GZMB
|
-0.084
|
5.49×10-3
|
*
|
-0.03822992
|
0.22850157
|
|
Cor, R value of Spearman’s correlation; None, correlation without adjustment. Purity, correlation adjusted by purity. *P < .01; **P < .001;***P < .0001. Abbreviations: TAM, tumour-correlated macrophage; Tfh, follicular helper T cell; Th, T helper cell; Treg, regulatory T cell.
|
In addition, the protein expression level of eIF4E can be discovered by using clinical samples from HPA database. The immunohistochemical images showed that eIF4E shows moderate staining in breast cancer (Figure 6B). At the same time, we verified the expression level of eIF4E significantly related immune cell markers in the same breast cancer patients, including TAM markers (IL10), M2 macrophage markers (CD163), Th1 (STAT1), Th2 (GATA3, STAT6), Th17 (STAT3) and Treg markers (STAT5B), in which GATA3, STAT3 and STAT5B were moderately stained and the others were weakly positive. The difference of expression of immune markers in tumor tissues of patients with breast cancer was further discussed.
3.5 Analysis of co-expression genes of eIF4E in breast cancer
In order to figure out the biological significance of eIF4E in BRCA, the functional module of LinkedOmics was used to check the co-expression pattern of eIF4E in the BRCA cohort. As showed in figure 7A, 5315 genes (dark red dots) were significantly positively correlated with eIF4E while 8395 genes (dark green dots) were negatively correlated. The heat map showed the first 50 important genes positively and negatively correlated with PRPF3 (Figure.7B and C), of which UBE2D3 ubiquitin binding enzyme had the highest positive correlation (r =0.671112, P =5.68E-144). The co-expressed genes were described in detail in the supplementary table2.
Gene ontology (GO) terminology annotations made through gene set enrichment analysis (GSEA) showed that genes co-expressed by eIF4E were mainly involved in chromosome segregation, RNA localization and DNA replication while extracellular structure organization, human immune response and protein localization to endoplasmic reticulum were inhibited(Figure7D, supplementary table3). KEGG enrichment showed that it was mainly concentrated in ubiquitin-mediated proteolysis, RNA transport, cell cycle and other signal pathways, while ribosome, glycosaminoglycan biosynthesis, cell adhesion molecules and other signal pathways were inhibited (Figure7E, supplementary table4).
In addition, co-expressed network of protein-protein interactions by Differential Net was constructed based on breast-specific data collected from eIF4E database (Figure.8A, supplementary table5). The top three central genes were CUL3, heat shock protein 90αA1 (HSP90AA1) and YWHAZ. CUL3 is the core component of BCR (BTB-CUL3-RBX1) E3 ubiquitin protein ligase complex. The ubiquitin ligase complex mediates the ubiquitination of the target protein and subsequent proteasome degradation (18); the ubiquitin ligase complex BCR (KLHL25) participates in translation homeostasis by mediating ubiquitin and hypo-phosphorylated eIF4EBP1 (4E-BP1) degradation (19). Extracellular heat shock protein 90α (HSP90AA1) has been widely reported promoting tumor cell movement and tumor metastasis in many tumors. It has been observed that extracellular heat shock protein 90α can promote EMT and the migration of breast cancer cells in breast cancer (20). YWHAZ binds and stabilizes key proteins involved in signal transduction, cell proliferation and apoptosis (21). Studies have further shown that YWHAZ is involved in drug resistance in breast cancer (22).
Finally, the TF (transcription factor)-mi RNA regulatory interaction of eIF4E co-expressed genes was constructed based on RegNetwork database (Figure.8B, supplementary table6). The top three TF were upstream stimulating factor1 (USF1), CCCTC binding factor (CTCF) and transcription factor YY1. USF1 related studies have shown that USF1 can transcriptionally up-regulate the expression of FAK in lung cancer, thus activating the FAK signal pathway and promoting cell migration (23). USF1 is involved in the transcription of many proteins and plays an important role as a regulator in many diseases, including tumors(24).Studies have shown that CTCF expression is involved in tumorigenesis(25) and can be used as a transcription factor, to control gene expression by binding to the transcriptional initiation site(TSSs) of many genes(26).Some studies have shown that the binding and overexpression of transcription factor YY1 with BRCA1 promoter inhibits the proliferation and focus formation of nude mice cells and inhibits the growth of MDA-MB-231 tumor. In addition, tissue microarray detected that there was a positive correlation between the expression of YY1 and BRCA1 in human breast cancer (27, 28).
3.6 Cross analysis of eIF4E co-expression genes and immune marker genes
We showed that eIF4E co-expression gene GO analysis was involved in human immune-related biological processes, KEGG enrichment also showed that it was involved in the cell adhesion molecule pathway, which related to the expression of cytokines. In order to further explore the relationship between eIF4E co-expression genes and immune infiltration, we made a cross-analysis of 13710 co-expression genes and 30 immune marker genes significantly related to eIF4E. The results showed that there were 18 overlapping genes (Figure 9A). The interaction of these key genes was analyzed by Cytoscape software and GO analysis. The results showed that the key genes were mainly involved in the human immune response, the adaptive immune response, macrophage activation, extracellular structure organization and regulation of DNA metabolic process (Figure 9C). KEGG analysis showed that these key genes were mainly involved in inflammatory bowel disease (STAT4, HLA-DPB1, HLA-DRA, STAT6, FOXP3, HLA -DPA1), cell adhesion molecule pathways (CD8B, HLA-DPB1, PDCD1, HLA-DPA1), JAK-STAT signaling pathways (STAT4, STAT6), T cell receptor signaling pathway (PDCD1) (Figure 9B). These results suggested that eIF4E co-expression genes were involved in the regulation of tumor immunity and provided strong evidence that eIF4E was an important regulator of immune infiltration in breast cancer.