Overexpressed BSG related to the progression of lung adenocarcinoma with high-throughput data-mining, immunohistochemistry, in vitro validation and in silico investigation

Abstract

Purpose: Lung adenocarcinoma (LUAD) of non-small cell lung cancer (NSCLC) is a highly prevalent cancer with high mortality. The gene basigin (BSG) is strongly expressed in certain tumors. This study investigated the expression level of BSG in LUAD and its role in the poor prognosis of LUAD. Methods: The mRNA expression of BSG in LUAD was from GEO, Oncomine and TCGA database. Prognostic data were provided by SurvExpress. The expression of BSG protein was detected by immunohistochemistry (IHC). Cell function experiments of viability, proliferation and apoptosis assays were performed in A549. Clustering analysis of BSG co-expressed genes was generated by Gene Ontology (GO) Enrichment, KEGG pathway and protein-protein interaction (PPI) network. Results: BSG mRNA level was significantly over-expressed in LUAD (pooled SMD = 0.564, 0.448-0.681, P<0.001). The analysis in SurvExpress revealed that high expression of BSG indicated significantly poor prognosis (pooled HR = 1.20, 1.10-1.30, P<0.0001). IHC assay also showed that BSG protein expression was significantly up-regulated in LUAD (P<0.001), and positive BSG expression was notably associated with higher pathology grade (P = 0.041) and lymphatic metastasis (P = 0.014). Moreover, BSG can enhance the viability and proliferation ability (both P<0.001) and weaken cell apoptosis (P<0.001) in A549. The most enriched GO terms in the co-expressed genes of BSG were translation related enrichment. The KEGG pathway showed that these genes were markedly involved in Ribosome pathways. Conclusion: Up-regulated BSG in LUAD is related to advanced progression and poor prognosis by influencing cell viability, proliferation and apoptosis. BSG could be a potential biomarker for the targeted therapy of LUAD.

Introduction

It has been well-understood that lung cancer, with high morbidity and mortality in humans, ranks as the most threatening cancer in a suburban area of China [1], and even worldwide [2]. Approximately 80%-85% of lung cancer is determined to be non-small cell lung cancer (NSCLC), which has a poor prognosis. The curative effect and prognosis of the present treatments for NSCLC, which consist of surgical resection, radiofrequency ablation, radiotherapy, platinum-based chemotherapies and targeted therapy, are unsatisfactory [3,4]. Researchers are studying the invasive and metastatic mechanism of NSCLC at various molecular levels and signal pathways [5,6]. In this study, we investigated the related molecular mechanism of lung adenocarcinoma (LUAD), which belongs to NSCLC.

Increasing evidence shows that basigin (BSG, also named TCSF, CD147, EMMPRIN) could stimulate para-cancerous fibroblast secreting matrix metalloproteinases (MMP) [7]. The increased concentration and viability of MMP can cause excessive degradation of extracellular matrix and basement membrane of blood vessels. Consequently, cancer cells can easily penetrate the basement membrane of blood vessels and extracellular matrix causing invasion and distant metastasis [8].

Large-scale studies reported that BSG was overexpressed in various human cancers, such as hepatocellular carcinoma [9], colon cancer [10] and breast cancer [11]. Moreover, the role of BSG in NSCLC [12] had been examined to regulate MMP secretion and influence tumor invasion and metastasis. Additionally, evidence showed that BSG may also induce chemoresistance in cancer cells by promoting the synthesis of hyaluronan [13].

In this study, we made a deep investigation into BSG from high-throughput data-mining, clinical immunohistochemistry verification, in vitro validation and in silico investigation to have a comprehensive understanding of BSG.

Material and methods

Data source and selection criteria

Public high-throughput RNA-sequencing and microarray data were acquired from Gene Expression Omnibus (GEO) of National Center of Biotechnology Information (NCBI) database (https://www.ncbi.nlm.nih.gov/geo/), Oncomine database (https://www.oncomine.org/resource/main.html), and The Cancer Genome Atlas (TCGA) database (https://cancergenome.nih.gov/) [14]. SurvExpress database (http://bioinformatica.mty.itesm.mx:8080/Biomatec/SurvivaX.jsp) was used to explore the survival data of BSG in lung cancer tissues, which included 23 available databases [15]. The keywords searched in GEO were (BSG OR Basigin OR CD147 OR “Extracellular Matrix Metalloproteinase Inducer” OR EMMPRIN OR “M6 activation antigen” OR “collagenase stimulatory factor protein” OR HAb18G OR “human leukocyte activation antigen M6” OR “5F7 protein” or “CD147 Antigen”). In the GEO dataset search, GEO series, which met the following criteria were selected for our study: (1) Samples contained LUAD group and noncancerous control group. (2) Samples were from clinical patients. (3) Series detected expression profiling by RT-qPCR/array/high-throughput sequencing data.

In-house validation with immunohistochemistry

In this study, a total of 127 LUAD and 30 normal lung tissues were included in a tissue microarray. All of the patient tissues were collected from the Pathology Department, First Affiliated Hospital of Guangxi Medical University, P. R. China (January 2010 to February 2014). The research was authorized by the Ethical Committee of the First Affiliated Hospital of Guangxi Medical University and the usage of the tissue study had consent from clinical doctors and patients. All the clinicopathological features were obtained to analyze the relationship to BSG expression.

Immunohistochemistry and scoring criteria

The expression of BSG protein was detected in the aforementioned LUAD patients by IHC, which was performed on formalin-fixed and paraffin-embedded tissue array sections. After being deparaffinized and boiled in sodium citrate buffer (0.01 mol/L, pH = 6.0) in microwave for ten minutes, tissue sections were applied the anti-CD147 (BSG) antibody [EPR4052] (1:250 dilution, Abcam, USA), stained by avidin-biotin-peroxidase method and counterstained with hematoxylin.

Every sample was scored by two independent pathologists (Gang Chen and Kang-lai Wei). Five visions were randomly captured and assessed by the percentage and intensity of the stained area. The positive BSG area <10%, 10-25%, 25-50%, 50-75% and 75-100% graded 0, 1, 2, 3 and 4 point, respectively. The staining intensity weakly, moderately, strongly graded 1, 2 and 3 point, respectively. If the added two scores were greater than 2, the tissue sample was regarded as BSG positive; otherwise, it was negative [16].

Cell culture and establishment of siRNA infected cell line

The human LUAD cell line A549 was cultured at Roswell Park Memorial Institute (RPMI); 1640 medium (GIBCO, Los Angeles, CA, USA), which contained 10% fetal bovine serum (FBS) and 100 U/mL penicillin-streptomycin sulfate. The culture dishes were placed in the incubator with a humidified 5% CO₂ 37°C environment. Lipofectamine 2000 reagent (Invitrogen, Carlsbad, CA, USA) was applied to the cell transfection. Cells were transfected according to the manufacturer’s protocol. The siRNA targeting human BSG with the sequence of 5’-GGUCAGAGCUACACAUUGA-3’ and the siRNA used as negative control were purchased from Ambion, Life Technologies Europe B.V., Ghent, Belgium.

Cell proliferation trial

The cell proliferation trial was conducted by Promega MTS kit (G3580, Promega, USA). The cell line A549 was prepared in a 96-well plate. Next, reagent was added into the 96-well plate in the quantity of 20 μl per well. Two hour later, the absorbance fluorescence values were measured at 490 nm [17].

Cell viability assay

The cell viability assay was conducted by CellTiter-Blue Cell Viability Assay kit (G8080, Promega, USA). The cell line A549 was prepared in a 96-well plate. Next, reagent was added into the 96-well plate in the quantity of 20 μl per well. Two hours later, the absorbance fluorescence values were measured at 560/590 nm.

To obtain a reliable result of cell viability, Hoechst 33342 and Propidium Iodide (PI) (Sigma, USA) double fluorescent staining assay were used to stain the cells. The stained cells were observed under microscope (200×). Normal cells showed weak red and weak blue fluorescence. Apoptotic cells showed strong blue and weak red fluorescence. Necrotic cells showed strong blue and red fluorescence [18].

Cell apoptosis experiment

The Apo-ONE Homogeneous Caspase-3/7 Assay (G7790, Promega, USA) was applied to cell apoptosis experiment. The cell line A549 was prepared in a 96-well plate. The reagent was mixed into the 96-well plate containing blank, control and samples. One hour later, the absorbance fluorescence values were measured with the excitation wavelength for detection of 499 nm, and the emission maximum at a wavelength of 521 nm.

To get a reliable result of cell apoptosis, Hoechst 33342 and PI double fluorescent staining assay was also adopted to stain the cells as above [19].

Clustering analysis of BSG co-expressed genes

To analyze the potential pathways related to BSG, co-expressed genes associated with BSG were obtained from Multi Experiment Matrix (MEM) database (http://biit.cs.ut.ee/mem/index.cgi) [20,21] and cBioPortal for Cancer Genomics (cBioPortal) database (http://www.cbioportal.org). Venny 2.1.0 (By Juan Carlos Oliveros) tool (http://bioinfogp.cnb.csic.es/tools/venny/index.html) was applied to achieve the intersection co-expression genes of the two databases. Next, the obtained BSG co-expressed genes were analyzed by DAVID Bioinformatics Resources 6.8 database (https://david.ncifcrf.gov/) for Gene Ontology (GO) clustering analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis. STRING: Functional protein association networks 10.5 (https://string-db.org/) was the approach for protein-protein interaction (PPI) analysis [22].

Statistical analysis

All the original data of high-throughput expression databases were transformed by exponential log2. The mean value of BSG expression and standard deviation (SD) were calculated by SPSS 22.0 (IBM, New York, USA). Also, the standard mean difference (SMD) and 95% confidence interval (CI) were calculated by Stata 12.0 (StataCorp, College Station, TX, USA).

Z-test was performed to evaluate the significance of pooled estimates with two-sided p-values and was considered significant when P<0.05. Q-test chi-square and the Higgins I-squared were estimated to assess heterogeneity across studies [23]. I-squared values of 75%, 50% and 25% manifested high, moderate and low heterogeneity, respectively. A fixed-effects model (Mantel-Haenszel method) was used to estimate the overall pooled results, if P<0.1 in Q-test. Otherwise, a random-effect model (DerSimonian-Laird method) would be applied. Sensitivity analysis was adopted by the means of omitting one study at a time to evaluate the reliability of the pooled estimates. Publication bias was assessed by the symmetry of funnel plot visually and by Egger’s and Begg’s linear regression statistically. All statistical results were obtained by Stata 12.0 (StataCorp, College Station, TX, USA).

Results

Expression of BSG in public high-throughput datasets

After searching from the GEO datasets, eighteen BSG expression datasets obtained from GEO or Oncomine and the LUAD data from TCGA database were included into this study (Table 1). We calculated the expression level of BSG in LUAD from the eighteen high-throughput datasets of RNA-sequencing and microarray data. The pooled SMD and 95% CI were conducted by a fix-effects model when we deleted eight datasets to make the heterogeneity I-squared <50% among the eighteen studies (I-squared = 48.6%, P = 0.011). The expression of BSG in LUAD was significantly higher than the control group (pooled SMD = 0.564, 95% CI: 0.448-0.681, P<0.001, Figure 1A). The expression of BSG in every study of GEO datasets is shown in Figure 2. The ROC curves of BSG expression in GEO are shown in Figure 3.

Table 1.

Main characteristics and results of the eligible studies of GEO, Oncomine and TCGA database

Data source	Series	Author	Country	Platform	Source	Total sample	LUAD group			Control group			t	p-value
							Number	Mean	SD	Number	Mean	SD
GEO	GSE101929	Mitchell	USA	GPL570	tissue	66	32	9.9831	0.6555	34	9.7959	0.5319	1.277	0.206
GEO	GSE89039	Zhang	China	GPL17077	tissue	16	8	2.0661	0.30253	8	1.8223	0.09485	2.175	0.06
GEO	GSE75037	Girard	USA	GPL6884	tissue	166	83	12.328	0.8279	83	11.5692	0.51708	7.082	<0.001
GEO	GSE69732	Luo	Taiwan, China	GPL14951	blood	6	3	7.8316	0.59863	3	8.177	0.7698	-0.614	0.573
GEO, Oncomine	GSE68571	Heiskanen	USA	GPL80	tissue	96	86	10.6057	0.70797	10	10.3465	0.27471	1.143	0.256
GEO	GSE33532	Meister	Germany	GPL570	tissue	100	40	8.5114	0.64197	20	8.1396	0.46914	2.547	0.014
GEO	GSE21933	Wang	Taiwan, China	GPL6254	tissue	42	11	10.5396	0.73921	21	9.8929	0.86131	2.112	0.043
GEO	GSE29249	Ma	China	GPL10558	tissue	12	3	9.9916	0.41926	3	10.1748	0.14972	-0.713	0.515
GEO, Oncomine	GSE32863	Laird-Offringa	USA	GPL6884	tissue	116	58	12.2309	0.90385	58	11.443	0.50775	5.788	<0.001
GEO, Oncomine	GSE31210	Kohno	Japan	GPL570	tissue	246	226	8.761	0.58791	20	8.5503	0.27429	2.897	0.006
GEO	GSE20189	Rotunno	USA	GPL571	blood	162	73	9.5296	0.80979	81	9.1888	0.7948	2.633	0.009
GEO	GSE10799	Streichert	Germany	GPL570	tissue	19	16	7.7469	0.65006	3	7.3571	0.23229	1.006	0.328
GEO	GSE11117	Budach	Switzerland	GPL6650	tissue	56	13	8.7007	0.74643	15	8.2783	0.7944	1.443	0.161
Oncomine	GSE2514	Stearman	USA	GPL8300	tissue	39	20	11.8006	0.27324	19	11.781	0.16047	0.275	0.785
GEO	GSE1987	Dehan	Israel	GPL91	tissue	37	7	11.4312	0.70249	7	11.7023	0.64842	-0.75	0.468
GEO, Oncomine	GSE10072	Lee	USA	GPL96	tissue	107	58	8.956	0.4501	49	8.7295	0.48385	2.505	0.014
GEO, Oncomine	GSE7670	Su	Taiwan, China	GPL96	tissue	66	31	9.8428	0.73624	27	9.3547	0.4858	3.014	0.004
TCGA	TCGA-LUAD	TCGA	USA	None	tissue	562	503	13.375	0.605	59	13.174	0.321	4.06	<0.001

Figure 1.

The expression of BSG in LUAD from GEO, Oncomine and TCGA database. A. Forest plot. B. Sensitivity analysis plot. C. The Begg’s funnel plot of the publication bias. Abbreviations: LUAD, lung adenocarcinoma; CI, confidence interval; SMD, standard mean difference.

Figure 2.

Differential expression of BSG in cancer and non-cancerous tissues from GEO datasets. *P<0.05.

Figure 3.

ROC curve of BSG expression in LUAD. *P<0.05. Abbreviations: LUAD, lung adenocarcinoma; AUC, area under the curve; ROC, receive operator characteristic curve.

The sensitivity analysis was conducted to verify that the pooled SMD was stable (Figure 1B). The Begg’s and Egger’s tests showed the publication bias was not significant when the p-values were greater than 0.05 (Begg’s P = 0.544, Egger’s P = 0.303, Figure 1C).

Prognostic role of BSG in lung cancer

SurvExpress database showed that the expression of BSG in lung cancer patients indicated significantly poor prognosis (lung cancer vs non-cancerous tissue, HR = 1.20, 95% CI: 1.10-1.30, P<0.0001) (Figure 4A). The Begg’s funnel plot and Egger’s publication bias plot are shown in Figure 4B, 4C.

Figure 4.

The association between BSG expression and OS in lung cancer from SurvExpress. A. Forest plot. B. Begg’s funnel plot. C. Egger’s publication bias plot. Abbreviations: OS, overall survival; HR, hazard ratio; CI, confidence interval.

Expression of BSG in patient tissues by immunohistochemistry

As shown in Table 2, the expression status of BSG was significantly different between LUAD and para-carcinoma tissue (P<0.001) by IHC assay. For the clinicopathological parameters, BSG expression was significantly associated with pathology grade (P = 0.041) and lymphatic metastasis (P = 0.014). The Hematoxylin and Eosin (HE) staining and IHC images of BSG-stained LUAD are shown in Figure 5A-H. The staining of BSG protein was primarily expressed in cytomembrane.

Table 2.

Relationship of BSG expression and clinicopathological parameters by immunohistochemistry

	Total number	BSG (-)	BSG (+)	Z score	P-value
Tissue
LUAD	127	64	63	-3.601	<0.001*
pT	30	26	4
Gender
Male	71	36	35	-0.629	0.529
Female	34	15	19
Age
≤60	70	37	33	-1.237	0.216
>60	35	14	21
Pathology grade
I	8	5	3	6.368a	0.041*
II	45	24	21
III	33	9	24
T stage
T1-T2	106	52	54	-0.674	0.5
T3-T4	21	12	9
Distant metastasis
No	96	49	47	-1.646	0.1
Yes	9	2	7
Lymphatic metastasis
No	82	48	34	-2.468	0.014*
Yes	45	16	29
Tumor size
≤7 cm	105	51	54	-0.894	0.371
>7 cm	22	13	9

^*P<0.05.

Figure 5.

H-E staining and IHC staining for BSG expression in LUAD and detection of the effect of BSG on LUAD cell line A549 viability, proliferation and apoptosis ability. A. HE staining of LUAD (×100); B. IHC of LUAD (×100); C. HE staining of LUAD (×400); D. IHC of LUAD (×400); E. HE staining of LUAD (×100); F. IHC of LUAD (×100); G. HE staining of LUAD (×400); H. IHC of LUAD (×400). I. Cell viability detected by CellTiter-Blue cell viability kits. J. Cell proliferation detected by MTS kit. K. Cell viability detected by Hoechst 33342 and PI double fluorescent staining assay. L. Cell apoptosis detected by Caspase-3/7 assay. M. Cell apoptosis detected by Hoechst 33342 and PI double fluorescent staining assay. Abbreviations: LUAD, lung adenocarcinoma; H-E, Hematoxylin-eosin; IHC, immunohistochemical; PI, Propidium Iodide.

Cell viability and proliferation inhibited by BSG siRNA

The LUAD cell line A549 was adopted to undertake the cell function experiments. Cell viability trial by CellTiter-Blue Cell Viability Assay kit (Figure 5I) and by Hoechst 33342 and PI double fluorescent staining assay (Figure 5K) both showed that the viability of cells decreased significantly on the tenth day after BSG siRNA transfection and the down-regulated expression of BSG (both P<0.001). In the cell proliferation assay, the result showed that when BSG siRNA was transfected, the proliferation ability of A549 was significantly weakened (P<0.001, Figure 5J).

Cell apoptosis promoted by BSG siRNA

The Apo-ONE Homogeneous Caspase-3/7 Assay (Figure 5L) and Hoechst 33342 and PI double fluorescent staining assay (Figure 5M) both presented that cell apoptosis was remarkably up-regulated after BSG transfection (P<0.001) on the tenth day.

Potential pathways related to BSG

The 450 common elements were from BSG co-expressed genes selected between the Affymetrix GeneChip Human Genoma U133 Plus 2.0 collection (containing 1794 datasets) by MEM and the mRNA Expression z-Scores (RNA Seq V2 RSEM) profile by cBioPortal. From the DAVID Bioinformatics Resources, the GO cluster and KEGG pathway are listed in Tables 3, 4. The most significant Biological process (BP) categories were translation and translational elongation. The most significant Cellular component (CC) categories were ribonucleoprotein complex, ribosome and ribosomal subunit. The most significant Molecular function (MF) categories were structural constituent of ribosome, RNA binding and structural molecule activity. KEGG pathway showed that the most significant pathways were Ribosome and Oxidative phosphorylation. Within the 450 selected genes, the hub genes were acquired from STRING. Next, we placed the top 100 hub genes into STRING again and generated the PPI network (Figure 6).

Table 3.

GO functional annotations for most significantly co-expressed genes of BSG

GO ID	GO term	Count (%)	P-Value	Genes symbol
Biological process
GO:0006412	translation	47 (11.41)	1.12E-22	RPL18, NACA, MRPL41, RPL14, MRPS12, MRPS11, RPL35, RPS6KB2, RPL36, EIF5A,etc
GO:0006414	translational elongation	26 (6.31)	6.05E-19	RPL18, RPL14, RPL35, RPL36, RPLP2, EEF2, RPS5, RPL28, RPL29, RPS3, etc
GO:0022900	electron transport chain	20 (4.85)	2.20E-11	NDUFA2, NDUFA3, UQCRC1, NDUFB10, NDUFB7, NDUFA7, NDUFA13, NDUFC1, UQCRQ, NDUFA11, etc
GO:0006091	generation of precursor metabolites and energy	30 (7.28)	3.83E-10	ATP5D, ALDOA, UQCRC1, NDUFB7, ATP5G2, UQCRQ, ATP6V0B, NDUFS7, CYB561D2, NDUFS5, etc
GO:0006119	oxidative phosphorylation	17 (4.13)	1.19E-09	ATP5D, NDUFA2, NDUFA3, NDUFB10, UQCRC1, NDUFB7, NDUFA7, ATP5G2, NDUFC1, ATP6V0B, etc
GO:0006396	RNA processing	39 (9.47)	2.83E-09	POLR2F, POLR2E, PRPF4B, RPL14, POLR2L, PUSL1, POLR2I, LSM7, TRMT1, TRPT1, etc
GO:0042775	mitochondrial ATP synthesis coupled electron transport	13 (3.16)	5.89E-09	NDUFA2, NDUFA3, NDUFB10, UQCRC1, NDUFB7, NDUFA7, NDUFC1, NDUFS7, NDUFS5, UQCR11, etc
GO:0042773	ATP synthesis coupled electron transport	13 (3.16)	5.89E-09	NDUFA2, NDUFA3, NDUFB10, UQCRC1, NDUFB7, NDUFA7, NDUFC1, NDUFS7, NDUFS5, UQCR11, etc
GO:0022904	respiratory electron transport chain	13 (3.16)	2.92E-08	NDUFA2, NDUFA3, NDUFB10, UQCRC1, NDUFB7, NDUFA7, NDUFC1, NDUFS7, NDUFS5, UQCR11, etc
GO:0034470	ncRNA processing	20 (4.85)	1.05E-07	RPL14, EXOSC4, PUSL1, TRPT1, TRMT1, RRP9, EXOSC1, QTRT1, NOP10, URM1, etc
Cellular component
GO:0030529	ribonucleoprotein complex	68 (16.50)	2.03E-30	RPL18, SRP14, MRPS34, MRPL41, PRPF4B, RPL14, LSM7, RPLP2, RPS6KB2, RPS3, etc
GO:0005840	ribosome	44 (10.68)	3.18E-27	RPL18, MRPS34, MRPL41, RPL14, MRPS12, MRPS11, RPL35, RPS6KB2, RPL36, RPLP2, etc
GO:0033279	ribosomal subunit	34 (8.25)	7.89E-25	RPL18, MRPL41, RPL14, MRPS12, MRPS11, RPL35, RPLP2, RPL36, RPS3, MRPL12, etc
GO:0005739	mitochondrion	79 (19.17)	1.17E-18	ATP5D, MRPS34, TSPO, MRPL41, PTGES2, UQCRC1, PRDX5, ROMO1, COX5A, C14ORF2, etc
GO:0022626	cytosolic ribosome	23 (5.58)	1.88E-17	RPL18, RPL14, RPL35, RPLP2, RPL36, RPS5, RPL28, RPS3, RPL29, RPS19, etc
GO:0015934	large ribosomal subunit	20 (4.85)	1.37E-15	RPL18, MRPL41, RPL14, RPL35, RPLP2, RPL36, RPL28, RPL29, MRPL23, MRPL12, etc
GO:0044429	mitochondrial part	52 (12.62)	3.61E-15	ATP5D, TSPO, UQCRC1, MRPL41, COX5A, UQCRQ, COX5B, NDUFAF3, NDUFS7, LONP1, etc
GO:0044445	cytosolic part	27 (6.55)	4.53E-15	RPL18, RPL14, PIK3C2A, RPL35, RPL36, PRDX5, RPLP2, RPS5, RPL28, RPL29, etc
GO:0005743	mitochondrial inner membrane	33 (8.01)	4.25E-12	ATP5D, UQCRC1, NDUFB7, MTX1, ATP5G2, COX5A, UQCRQ, COX5B, NDUFAF3, NDUFS7, etc
GO:0019866	organelle inner membrane	34 (8.25)	6.11E-12	ATP5D, UQCRC1, NDUFB7, MTX1, ATP5G2, COX5A, UQCRQ, COX5B, NDUFAF3, NDUFS7, etc
Molecular function
GO:0003735	structural constituent of ribosome	39 (9.47)	1.29E-27	RPL18, MRPL41, RPL14, MRPS12, MRPS11, RPL35, RPL36, RPLP2, RPS3, MRPL12, etc
GO:0003723	RNA binding	46 (11.17)	3.75E-10	RPL18, SRP14, RPL14, PUSL1, LSM7, RPL35, EIF5A, RPLP2, TRMT1, CALR, etc
GO:0005198	structural molecule activity	41 (9.95)	3.32E-09	RPL18, MRPL41, CLTB, RPL14, MRPS12, MRPS11, RPL35, RPL36, RPLP2, RPS3, etc
GO:0016651	oxidoreductase activity, acting on NADH or NADPH	15 (3.64)	4.27E-09	NDUFA2, NDUFA3, NDUFB10, NDUFB7, NDUFA7, NDUFA13, CYB5RL, NDUFC1, ECSIT, NDUFS7, etc
GO:0008137	NADH dehydrogenase (ubiquinone) activity	11 (2.67)	2.92E-08	NDUFS7, NDUFS5, NDUFA2, NDUFA3, NDUFB10, NDUFB7, NDUFS8, NDUFA7, NDUFA13, NDUFC1, etc
GO:0050136	NADH dehydrogenase (quinone) activity	11 (2.67)	2.92E-08	NDUFS7, NDUFS5, NDUFA2, NDUFA3, NDUFB10, NDUFB7, NDUFS8, NDUFA7, NDUFA13, NDUFC1, etc
GO:0003954	NADH dehydrogenase activity	11 (2.67)	2.92E-08	NDUFS7, NDUFS5, NDUFA2, NDUFA3, NDUFB10, NDUFB7, NDUFS8, NDUFA7, NDUFA13, NDUFC1, etc
GO:0015078	hydrogen ion transmembrane transporter activity	14 (3.40)	1.05E-07	ATP5D, UQCRC1, UQCR11, UQCRH, COX8A, COX6B1, COX4I1, COX6A1, ATP5G2, COX5A, etc
GO:0016655	oxidoreductase activity, acting on NADH or NADPH, quinone or similar compound as acceptor	11 (2.67)	1.11E-07	NDUFS7, NDUFS5, NDUFA2, NDUFA3, NDUFB10, NDUFB7, NDUFS8, NDUFA7, NDUFA13, NDUFC1, etc
GO:0015077	monovalent inorganic cation transmembrane transporter activity	14 (3.40)	5.88E-07	ATP5D, UQCRC1, UQCR11, UQCRH, COX8A, COX6B1, COX4I1, COX6A1, ATP5G2, COX5A, etc

Table 4.

KEGG pathway enrichment analysis of BSG co-expressed genes

KEGG ID	KEGG Term	Count (%)	P-Value	Genes symbol
hsa03010	Ribosome	23 (5.58)	2.43E-15	RPL18, RPL14, RPL35, RPLP2, RPL36, RPS5, RPL28, RPS3, RPL29, RPS19, etc
hsa00190	Oxidative phosphorylation	26 (6.31)	2.52E-14	ATP5D, UQCRC1, NDUFB7, ATP5G2, COX5A, COX5B, UQCRQ, ATP6V0B, NDUFS7, NDUFS5, etc
hsa05016	Huntington’s disease	29 (7.04)	1.66E-13	ATP5D, POLR2F, UQCRC1, CLTB, POLR2E, POLR2L, NDUFB7, AP2S1, POLR2I, ATP5G2, etc
hsa05012	Parkinson’s disease	22 (5.34)	8.69E-11	ATP5D, NDUFA2, NDUFA3, UQCRC1, NDUFB10, NDUFB7, NDUFA7, COX8A, COX4I1, ATP5G2, etc
hsa05010	Alzheimer’s disease	24 (5.83)	2.40E-10	ATP5D, NDUFA2, NDUFA3, UQCRC1, NDUFB10, NDUFB7, NDUFA7, COX8A, COX4I1, ATP5G2, etc
hsa04260	Cardiac muscle contraction	10 (2.43)	4.48E-04	UQCRC1, UQCR11, UQCRH, COX8A, COX6B1, COX4I1, COX6A1, COX5A, UQCRQ, COX5B
hsa00240	Pyrimidine metabolism	10 (2.43)	0.00188043	NME4, ITPA, NME2, POLR2F, POLR2E, NME3, POLR2L, POLD2, POLR2I, ZNRD1
hsa03040	Spliceosome	11 (2.67)	0.00398349	DHX8, LSM7, PQBP1, SNRPB, SNRPA, LSM4, SF3B5, SF3A2, BUD31, NAA38, etc
hsa00230	Purine metabolism	12 (2.91)	0.00546642	NME4, ITPA, NME2, POLR2F, POLR2E, NME3, POLR2L, POLD2, POLR2I, ZNRD1, etc
hsa03020	RNA polymerase	5 (1.21)	0.00877957	POLR2F, POLR2E, POLR2L, POLR2I, ZNRD1
hsa00051	Fructose and mannose metabolism	4 (0.97)	0.0786479	ALDOA, GMPPA, TSTA3, PHPT1
hsa05110	Vibrio cholerae infection	5 (1.21)	0.08368437	SEC61B, ARF1, KDELR1, ATP6V0B, ATP6V1F
hsa03018	RNA degradation	5 (1.21)	0.08802319	EXOSC4, LSM7, LSM4, EXOSC1, NAA38

Figure 6.

PPI network of the top 100 BSG co-expressed hub genes. Abbreviations: PPI, protein-protein interaction.

Discussion

In this study, the considerable evidence from high throughput analysis and clinical study all represented that BSG was up-regulated in LUAD tissue. BSG participated in various biological processes, especially in processes related to tumor progression.

The meta-analysis of BSG expression data from GEO showed that BSG was significantly up-regulated in LUAD. To reduce heterogeneity, we excluded studies that contained negative numbers (GSE74706, GSE66759, GSE19188, GSE11969 and GSE1037) because the array author may have processed the data by other means, which was different from making the original data log2 transformed. We thought different processing methods were not suitable together for statistics. We also excluded the studies (GSE85841, GSE18549) whose LUAD or control group SD>1, and one study (GSE68465) whose SMD (95% CI) drifted too far from the invalid line. Finally, we obtained a significant result (pooled SMD = 0.564, 95% CI: 0.448-0.681, P<0.001) with an applicable heterogeneity (I-squared = 48.6%, P = 0.011).

Within our research, BSG co-expressed genes were mainly enriched in translation related GO terms and pathways. We could deduce that BSG is an important gene which participates in various proteins translation process biological mechanism. It could be a possible reason that the function of BSG is considerable in diverse biological process.

To date, the molecular mechanism of BSG has not been fully investigated. BSG could have functions in tumor cell biological characteristics of proliferation, metastasis and apoptosis via specific signal pathways in various malignant neoplasms. Studies of BSG were searched in pulmonary disease as follows (BSG can be called CD147 or EMMPRIN in the following). It has been reported that CD147 could form a complex with CD98 heavy chain (CD98hc) on cell plasma membrane in certain cancers. Also, this complex led to the proliferation of tumor cells and poor prognosis in NSCLC patients through the PI3K/Akt pathway [24]. One research group found that EMMPRIN had a positive regulation on the canonical Wnt/β-catenin signaling pathway, which could gain the ability for tumor cell proliferation, differentiation and metastasis after altering cell-cell interactions [25]. β-catenin/canonical Wnt signaling pathway was also reported to have a function in the chronic progressively fatal disease, idiopathic pulmonary fibrosis (IPF). TGF-β1 induced this pathway to release EMMPRIN, and overexpressed EMMPRIN may be relevant to the persistent fibro-proliferative state of IPF [26]. Regarding other diseases, CD147 could be blocked by a murine antibody HAb18 to inhibit hepatocellular carcinoma invasion and migration. Inducing the integrin signaling pathway, cHAb18 also suppressed cell motility by rearranging actin cytoskeleton, which made a new tumor therapy targeting CD147 [27]. Another group recently published that elevated copper levels promoted cancer cell metastasis as an extracellular signaling role, for the reason that Cu²⁺-mediated self-association of CD147 was able to activate PI3K/Akt signaling pathway, leading to up-regulation of the level of matrix metalloproteinase MMP-2 and MMP-14 in HCC cells [28]. Smad4 was found to be a new interaction copartner of CD147 and could inhibit the Smad4/p21 WAF1 signal, which promotes cell proliferation [29]. Physcion 8-O-β-glucopyranoside (PG) was proved to have an effect on EMMPRIN to regulate the ROS-miR-27a/ZBTB10-Sp1 transcription factor pathway in the osteosarcoma cell to induce cell apoptosis and cause cell cycle arrest [30]. Such compelling rationales of BSG mechanism in different pathways provided tremendous benefits for clinical anticancer therapies, especially terminal cancers.

Regarding the future research prospects, the function of BSG was unclear and clinical researchers should explore the association between BSG and various cancer types in the future. Finally, to provide evidence to the clinical therapy of LUAD, considerably more research on BSG alone or with other potential relevant molecules should be conducted.

Conclusions

Since the abnormal over-expressed BSG in LUAD is correlated with advanced progression and poor prognosis and could influence the tumor cell function at the molecular level, BSG could be a potential diagnostic and prognostic biomarker, as well as a potential candidate for target therapy in patients with LUAD.

Disclosure of conflict of interest

None.