Multi-omics analysis of expression profile and prognostic values of connexin family in LUAD

Abstract

Purpose

Our study first explored the expression differences and prognostic significance of Cx genes in pan-cancer and then focused on LUAD. Our objectives were to conducted a comprehensive analysis of the expression profile, prognostic significance, genetic alterations, potential biological functions and drug sensitivity of the Connexin gene family in LUAD.

Methods

We developed a comprehensive prognostic model for LUAD by combining risk scores with clinical features and created a nomogram to predict 1-, 3-, and 5-year overall survival. Using single-cell sequencing, we examined the expression and biological functions of the identified prognostic markers.

Results

Our risk model revealed that GJB2-5 play a critical role in the prognosis of LUAD patients, associated with many biological processes such as cell cycle, DNA damage, EMT, hypoxia, invasion, and metastasis. Furthermore, the connexin gene family is linked to transcriptional mechanisms such as the extracellular matrix (ECM), migration, mobility, angiogenesis, and the epithelial-mesenchymal transition (EMT) genetic program.

Conclusion

The risk model can be used as a potential prognostic factor for LUAD patients and may provide new insights into cancer treatment from perspective of the expression of Cx genes.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00432-023-05075-5.

Introduction

The most frequent histological subtype of lung cancer, known as lung adenocarcinoma (LUAD), is responsible for the majority of cancer deaths (Goldstraw et al. 2011; Siegel et al. 2018). Patients with metastasis have considerably poorer survival rates, notwithstanding the possibility of curative treatment for early-diagnosed cancer patients and improved long-term survival (Duggan et al. 2016).The Connexin gene family is a group of genes that encode connexin proteins. These proteins are integral membrane proteins that form gap junctions between adjacent cells, allowing for direct intercellular communication (Payton et al. 1969). In humans, the Connexin gene family consists of 21 genes that are located on different chromosomes (Goodenough and Paul 2009; Bai 2016). Each connexin protein has a distinct tissue distribution and expression pattern, and plays an important role in various physiological processes such as cell growth, differentiation, and tissue homeostasis (Aasen et al. 2016). Cells frequently exhibit the expression of multiple isoforms of connexin, which are capable of intermixing to yield a vast repertoire of intercellular channel compositions. These compositions encompass various types, such as homomeric, heteromeric, homotypic, and heterotypic channels, and are instrumental in selectively regulating the exchange of ions and metabolome constituents between cells (Goodenough and Paul 2009). Thus, the potential formation of heteromeric channels by compatible connexins gives rise to a broad range of intercellular communication, resulting in an extensive spectrum of functional diversity (Goodenough and Paul 2009; Laird et al. 2017; Laird and Lampe 2018). And increasing evidence suggests that connexins play important roles in the development and progression of tumors.

Loewenstein and Kanno published a seminal ex vivo study in 1966, which revealed that the electrical coupling observed in healthy hepatocytes was absent in liver tumor cells (Loewenstein and Kanno 1966). Moreover, the findings of this study led to the hypothesis that the loss of direct intercellular communication is often associated with the occurrence and progression of cancer. Further investigations have established a connection between this phenomenon and the presence of gap junctions consisting of connexin proteins (McNutt and Weinstein 1969; Johnson and Sheridan 1971). Despite extensive research supporting the view that connexins act as tumor suppressors, Recent research suggests that some tumor forms, including NSCLC, glioma and malignant melanoma in humans, may promote various stages of carcinogenesis through connected and unconnected signaling pathways (Aasen et al. 2019). For instance, connexins have been shown to enhance the migration and invasion of tumor cells (Friedl and Wolf 2003; Wolf et al. 2007), form heterotypic gap junctions between tumor cells and endothelial cells to facilitate intravasation and extravasation(Saito-Katsuragi et al. 2007), and foster metastatic growth while potentially promoting resistance to cancer treatments(Osswald et al. 2015). Prior research had indicated differential expression of the same Cx in normal lung tissue, small tumors (0.5–1.5 mm), and larger tumors (≥ 2.5 mm) (Udaka et al. 2007). Studies also had revealed that a decrease in Cx43 levels can delay the development of detectable tumors, while maintaining normal levels of Cx43 can inhibit the metastasis of breast tumors to the lungs (Plante et al. 2011). There is a substantial association between Cx26 expression and unfavorable clinical outcomes, including metastasis, in squamous NSCLC (Ito et al. 2006). However, there is currently a lack of analysis exploring the impact of the entire Cx gene family expression on LUAD survival prognosis and the underlying mechanisms, with a basis in bioinformatics.

In this study, we reviewed the expression data of the Connexin gene family in TCGA and GEO databases and investigated the expression differences and prognostic impact of Connexin genes in pan-cancer at first. Subsequently, we conducted a comprehensive analysis of the expression profile, prognostic significance, genetic alterations, potential biological functions and drug sensitivity of the Connexin gene family in LUAD. We analyzed the mRNA and protein expression levels of the Connexin gene family in LUAD and explored their co-expression, interactions, and functional clustering. We evaluated somatic mutations and DNA copy number variations in 197 patients and explored DNA methylation and CNV of the Connexin gene family. We constructed a prognostic model based on differentially expressed Connexin genes and integrated clinical information, and investigated the mechanisms at the single-cell and transcriptome levels. Altered expression and function of Connexins in tumors may have significant implications for tumor growth, metastasis, and treatment. Our research results will provide more information about the importance of the Connexin gene family and open up new avenues for exploring its overall role in predicting prognosis, optimizing immune therapy, and reducing drug resistance.

Materials and methods

Data retrieval and processing

Differential expression analysis and prognostic analysis of the Connexin family members were conducted on pan-cancer tissues and their corresponding normal tissues using the GSCALite database (http://bioinfo.life.hust.edu.cn/web/GSCALite/). The transcriptomic data, and Genetic mutation data the corresponding clinical data of patients with LUAD were downloaded from TCGA (http://cancergenome.nih.gov) database and the Gene Expression Omnibus (GEO; https://www.ncbi.nlm.nih.gov/geo/) database. In order to standardize the gene expression data across samples, FPKM values were converted to transcript per kilobase million (TPM) values and then normalized by Log2 conversion. Visualization of the expression differences of Connexin family members in TCGA-LUAD was performed using the R (version 4.2.1) and ggplot2 package.

Differential expression and co-expression of Cx

The differentially expressed Cx in TCGA-LUAD and normal lung tissues were detected using the R package limma (Ritchie et al. 2015) with |log2 fold change (FC)|> 1 and adjusted p < 0.05 as the cutoff threshold. The Spearman correlation method was employed to assess the pairwise correlation of Connexin family members in TPM formatted TCGA-LUAD data. And the results were visualized using the circlize package (Gu et al. 2014).

Functional enrichment analysis

To investigate the functional mechanisms of Cx in LUAD, we conducted Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis using the Cluster Profiler package (Yu et al. 2012) in R with all displayed items p < 0.05. The enriched GO terms, KEGG pathways and PPI network of the differentially expressed Cx in TCGA-LUAD were further annotated and visualized by Metascape (Zhou et al. 2019).

Construction of protein–protein interaction (PPI) networks

The String database (http://string-db.org) compiles a staggering amount of publicly accessible data on protein–protein interactions (Szklarczyk et al. 2019). The protein–protein interaction (PPI) network of the Cx was built using the String database. The protein–protein interaction (PPI) network of the Cx was built using the String database. The Cytoscape software was used for network visualization.

GSCA analysis

The GSCA database was utilized for the analysis of differential expression and pathway activity of Cx in LUAD, as well as for the analysis of GSVA and pathway activity(Liu et al. 2023). The CNV summary module exhibits a roster of the Copy Number Variations identified in LUAD for the genes that pertain to the Connexin family, acquired from the GSCA database. And the methylation of Cx and its mRNA expression are correlated, according to the GSCA database.

Establishment of the prognostic model

The survival package was utilized to conduct proportional hazards hypothesis testing and Cox regression analysis. Univariate and multivariate Cox regression was performed to create a Cx-related prognostic signature based on differential expression and calculate the risk score for each patient. Variables with p-values less than 0.1 univariate Cox regression analysis were included in the multivariate Cox regression.

The ggplot2 package was employed for visualizing the risk factor plot. The rms package was used to construct and visualize a nomogram-related model, as well as perform calibration analysis and visualization to examine the differences between predicted and actual probabilities at different time points.

Cell culture and quantitative real-time PCR (qRT-PCR)

Lung cancer cell lines (NCI-1975, HCC827, A549 and PC-9) and the normal human bronchial epithelial cell line Beas2B were purchased from the from Procell (Wuhan, China) and validated by short tandem repeat analysis. Beas2B, NCI-1975, HCC827 and PC-9were maintained in RPMI-1640 (Gibco, USA). A549 were grown in F12K Medium (Gibco, USA). Cells were incubated at 37 ℃ in a humidified atmosphere with 5% CO2.

The total RNA extraction was performed using RNA-easy Isolation Reagent (No. RC112-01, Vazyme, China). RT–qPCR was performed using a HiScript III 1st Strand cDNA Synthesis Kit (No. R312-01, Vazyme, China) and ChamQ™ Universal SYBR® qPCR Master Mix (No. Q712-02, Vazyme, China) according to the manufacturer’s instructions.

Single-cell data and correlation analysis

Using the Cancer Single-cell State Atlas (CancerSEA; http://biocc.hrbmu.edu.cn/CancerSEA/) database, we investigate the correlation of GJB2, GJB3, GJB4, and GJB5 with 14 cellular activities associated with primary tumors, such as cell cycle regulation, proliferation, and angiogenesis (Yuan et al. 2019). Additionally, we utilized the ggplot2 package for data management and visualization. Furthermore, we utilized the IMMUcan (https://immucanscdb.vital-it.ch) database to explore the expression of GJB2, GJB3, GJB4, and GJB5 at the single-cell level in LUAD (Camps et al. 2023).

Chemotherapeutic response prediction

Predicted the chemotherapeutic response for each sample based on the largest publicly available pharmacogenomics database [the Genomics of Drug Sensitivity in Cancer (GDSC), https://www.cancerrxgene.org/]. The R package “pRRophetic” was used to implement the prediction procedure (Geeleher et al. 2014a). Ridge regression was used to calculate the samples' half-maximal inhibitory concentration (IC50). The default values for each option were used.Gene expression duplicates were averaged as a mean value using the batch effect of combat and tissue type of all tissues (Geeleher et al. 2014b).

Statistical analysis

All statistical analyses were performed by version 4.2.1 of R software. The Wilcoxon test was used to compare two paired groups. The correlation analysis was evaluated with Spearman’s test. P < 0.05 was considered statistically significant except for special notes.

Results

mRNA expression and survival analysis of connexins

Differential expression of connexins and survival analyses across multiple cancer types were performed using the GSCALite database (Fig. 1A and B). In LUAD, we observed that members of the Cx family have an impact on expression differences and prognosis. We investigated the differential expression of connexins (P < 0.01, |logFC|> 1) in the TCGA-LUAD and GSE31210 database and present the results in Table 1. The results were visualized using violin plots (Fig. 1C) and heat maps (Fig. 1D). We observed a significant upregulation of GJB2, GJB3, GJB4, and GJB6, while GJA1, GJA4, GJA5, GJC1, GJC2, and GJD3 exhibited reduced expression in LUAD tumor tissue (Fig. 1C). Furthermore, the heat map demonstrated specific co-expression patterns of GJB7, GJB4, GJB3, GJB5, GJA1, GJA3, GJB2, and GJB6 (Fig. 1D).

Fig. 1.

Expression analysis of the connexin family A Differential expression analysis of connexin family members from GSCALite in pan-cancer tumor tissue and its matched normal tissue; B Prognostic analysis of connexin family members from GSCALite in pan-cancer; C Violin plots showing expression of connexin family members from TCGA LUAD dataset; D Heat map of expression differences between normal and tumor sample groups for TCGA LUAD dataset

Table 1.

The differential expression of connexins in the TCGA-LUAD and GSE31210 datasets

Gene	logFC of TCGA-LUAD	P value of TCGA-LUAD	logFC of GSE31210	P value of GSE31210
GJB1	1.69706212	2.11E-08	1.41861458	2.21E-04
GJB2	4.52220016	3.66E-68	3.49341617	7.04E-13
GJB3	3.21228092	6.40E-24	0.34157864	1.40E-01
GJB4	2.59342311	1.33E-05	0.37205161	4.11E-02
GJB5	2.34577491	4.96E-09	0.43730663	2.49E-01
GJB6	4.37780231	1.69E-18	3.21921623	4.55E-06
GJB7	0.26609811	5.46E-01	–	–
GJA1	– 1.2661907	5.38E-15	0.51526708	1.85E-03
GJA3	0.89722653	3.58E-01	– 1.39870790	4.81E-03
GJA4	– 1.4933045	5.31E-36	– 1.14057665	1.58E-14
GJA5	– 1.7187275	2.07E-39	0.40927892	5.74E-07
GJA8	– 0.0363332	9.94E-01	0.51526708	1.09E-01
GJA9	–	–	–	–
GJA10	0.05466762	9.88E-01	0.29906040	2.44E-01
GJC1	– 0.189096	2.32E-01	– 1.28176592	3.66E-12
GJC2	– 1.468692	8.42E-24	– 0.97161203	2.22E-05
GJC3	0.60879	1.13E-01	–	–
GJD2	0.02255	9.96E-01	0.28007499	1.41E-02
GJD3	– 0.876798	2.36E-20	– 0.75915056	1.25E-08
GJD4	– 0.036333	9.94E-01	0.12907635	4.26E-01

Co-expression, interaction and functional analysis of connexins

We performed a co-expression analysis on the connexin family members to explore their relationships (Supplement Tables 1 and 2). The results of the analysis revealed that GJB2, GJB3, GJB4, GJB5, and GJD3 had stronger correlations with other family members in LUAD's transcriptome (Fig. 2A). The GO analysis indicated that connexins were predominantly enriched in cell–cell junction assembly (BP), connexin complex (CC), and gap junction channel activity (MF) (Fig. 2B and Table S3). Moreover, the KEGG pathway analysis revealed that the differentially expressed connexins were primarily associated with arrhythmogenic right ventricular cardiomyopathy and gap junction (Fig. 2B and Table S3). We also performed an annotation of GO and KEGG pathway enrichment analysis, as well as PPI network analysis of differentially expressed connexins in LUAD using the “Metascape” website (Fig. 2C). We also used the STRING database to cluster and construct a network of connexins and their neighbor genes and drew a network diagram, which revealed that GJA1 was the most important hub gene in the network (Fig. 2D). Our findings revealed that GJB2, GJA5, and GJA1 played an active role in EMT, while GJB1 inhibited apoptosis. Additionally, GJC2, GJB1, and GJA5 played an inhibitory role in the cell cycle, and GJB1 inhibited EMT. Furthermore, GJB3 and GJB2 inhibited the PI3K-AKT pathway and TSC-mTOR pathway (Fig. 2E). The difference in pathway activity between the high and low GSVA score of the differentially expressed connexin groups showed an inhibition of the PI3K-AKT pathway and TSC-mTOR pathway (Fig. 2F).

Table 2.

Correlation between the expression of GJB2-5 and the clinical pathological characteristics of lung adenocarcinoma

Variable	Coefficient β	HR	Confidence interval	p value
GJB4	− 0.12443	0.883	0.686–1.136	0.3327
GJB5	0.3503	1.419	0.879–2.293	0.1523
GJB3	0.026643	1.027	0.856–1.232	0.7745
GJB2	0.14649	1.158	0.994–1.349	0.0601
Pathologic T stage
T1		Reference
T2	0.55639	1.744	0.813–3.742	0.1531
T3	1.3601	3.897	1.443–10.519	0.0073
T4	0.72172	2.058	0.609–6.958	0.2455
Pathologic M stage
M0		Reference
M1	0.46401	1.590	0.670–3.775	0.2928
Age
<  = 65		Reference
> 65	0.16292	1.177	0.702–1.974	0.5370
Gender
Female		Reference
Male	0.18754	1.206	0.719–2.024	0.4778
Location
Central lung		Reference
Peripheral lung	0.23642	1.267	0.711–2.256	0.4222

Fig. 2.

Functional Annotation of the connexin family A Chord diagram shows expression correlation between connexin family members; B Circle plots of GO terms and KEGG pathways among genes that belong to the connexin family; C Website analysis using Metascape. The same hue signifies the same cluster ID, and nodes with the same cluster ID are generally close to one another; D Protein–protein interaction (PPI) network; E Heat plot of the percentage of LUAD where the differentially expressed connexin family genes may affect pathway activity; F Heat plot of the correlations between GSVA score and pathway activity among LUAD

Genetic alterations variation and methylation of Cx in LUAD

We conducted a mutation analysis of the connexin family members, which was presented in a waterfall plot. The top 5 commonly mutated genes were GJA8, GJA10, GJA9, GJB4, and GJA1. Among them, GJA8 had the highest mutation rate with more than 15% (Fig. 3A). The most prevalent mutation types in LUAD were Amplification, Deep Deletion, and Missense Mutation, while the most frequent variant classification was SNP. Among all SNP mutations, C > A mutation had the highest frequency (Fig. 3B, C, and D). DNA methylation is positively connected with GJA1's mRNA expression, while it is negatively correlated with the majority of Cx, such as GJB1, GJB2, GJD3, GJC2, GJB5, GJB3, GJA4, GJB6, GJA5, GJC1, GJC3, GJA9, and GJD2 (Fig. 3E). The heterozygous/homozygous CNV (deletion/amplification) status of GJA3, GJB6, GJB2, GJB2, GJB7, GJA1 and GJA10 had a similar alteration trend, while GJA5, GJA8 and GJC2 had a similar alteration trend (Fig. 3F).

Fig. 3.

Mutation profile analysis of Cx by using the TCGA datasets A Oncoplot displaying the somatic landscape of connexin family members from TCGA-LUAD database; B Overlay of histograms showing frequency of connexins in LUAD by cBioPortal; C The SNV classes of Cx in LUAD; D The Transitions (Ti) and transversions (Tv) classification of the SNV of Cx in LUAD. E Correlations between DNA methylation and mRNA expression of Cx were determined by TCGA; F Alterations of CNV of Cx in LUAD were assessed from GSCA

The correlation analysis of Cx with several genetic programs of cancer progression in LUAD

We observed that most of the Cx, except for GJB1, are not only implicated in the EMT genetic program, but also in transcriptional programs that promote ECM production, migration, motility, and angiogenesis. Interestingly, GJB1 showed a negative correlation with these programs' related genes (Fig. 4A-D). GJB2, GJB3, GJB4, GJB5, GJB6, and GJA1 show positive correlation with the process of adherence junctions (Fig. 4E).

Fig. 4.

The correlation coefficients of Cx mRNA with different functional states genes in LUAD A–E The correlation coefficients of Cx mRNA with the genes of EMT, ECM and surface adhesion, cellular migration and motility, angiogenesis and adherens junctions in LUAD

Construction of a prognostic risk model and risk score validation

Univariate Cox regression analyses were conducted to investigate the association between the expression of Cx and LUAD prognosis. The findings revealed that 10 Cx displayed a significant correlation with the survival of patients with LUAD (Fig. 5A). Furthermore, a multivariate Cox regression analysis was conducted to establish a prognostic risk model utilizing four genes, namely GJB2, GJB3, GJB4, and GJB5 (Fig. 5B). The gene expression heatmaps and survival overviews for the high-risk and low-risk groups, respectively. And the expression levels of GJB2, GJB3, GJB4, and GJB5 were high in the high-risk group (Fig. 5C). The Nomograms that predict OS in LUAD patients are presented in Fig. 5D. The Prognosis Nomogram is based on a multivariable regression analysis and provides a column chart for predicting OS based on nine prognostic factors, namely, the expression levels of GJB2, GJB3, GJB4, and GJB5, pathologic TNM staging (with N stage existing and collinearity of other variables removed), age, sex, and location (Table 2). A higher total point score, which is derived from the sum of the assigned points for each element in the nomograms, was associated with a poorer prognosis. To assess the accuracy of the Nomograms, we conducted a calibration analysis. We observed that the survival rates at 1, 3, and 5 years were closely aligned with the ideal curve, indicating a minimal difference between the predicted probabilities of our Nomograms and the actual probabilities (Fig. 5E).

Fig. 5.

Prognostic value of the significant Cx in LUAD patients A Univariable and multivariable Cox regression analysis between Cx expression and OS; B The distribution of risk scores based on prognostic signatures; C The area under the time-dependent ROC curve was utilized to evaluate the prognostic efficacy of the risk score in predicting the 5-year overall survival; D A nomogram was constructed to predict the probabilities of overall survival for LUAD patients at 1, 3, and 5 years; E The nomogram's calibration for the Cox regression model and its fitting analysis to the actual situation were performed

The expression of GJB2-5 in TCGA-LUAD,GSE31210 datasets and different LUAD cell lines

The heatmaps respectively display the differential gene expression in TCGA and GSE31210 databases, as well as their correlation with risk score, survival/death events, and time (Fig. 6A and B). Their trend of results is consistent. We found that the survival probability between the high-risk and low-risk groups has statistical significance with a P-value < 0.001, and the low-risk group demonstrates a better prognosis (Fig. 6C). In the GSE31210 dataset, the low-risk group also demonstrates a better prognosis, and the survival curves between the high-risk and low-risk groups are more separated (Fig. 6D).

Fig. 6.

The relationship between consensus clustering and overall survival was examined in both TCGA and GSE31210 datasets A Heatmap showing sample clusters with distinct clinical outcomes, time and risk score in TCGA-LUAD; B Heatmap showing sample clusters with distinct clinical outcomes, time and risk score in GSE31210; C Kaplan–Meier overall survival (OS) curves using TCGA-LUAD data; D Kaplan–Meier overall survival (OS) curves using GSE31210 datasets; E–H The mRNA expression levels of GJB2 (E), GJB3 (F), GJB4 (G) andGJB5 (H) in different cell lines (Beas2B, NCI-1975, HCC827, A549 and PC-9) were measured by RT-qPCR. Results were normalized to reference gene GAPDH. Data are shown as the mean ± SEM, two-tailed unpaired t test was used for statistical calculation for each marker, n = 3 independent experiments. (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.001; ns, not significant)

To validate the outcomes of our data analysis, we isolated total RNA from various LUAD cell lines (NCI-1975, HCC827, A549 and PC-9) and the normal human bronchial epithelial cell line Beas2B, and quantified the mRNA expression levels of GJB2, GJB3, GJB4, and GJB5. RT–qPCR assays revealed that GJB2, GJB3, GJB4, and GJB5 are highly expressed in LUAD cell lines. (Fig. 6E-H). Table 3 summarizes the details on primers which were used for this study.

Table 3.

Primer sequences of genes used for qRT-PCR

Human_GJB2	Forward: 5′-TCGCATTATGATCCTCGTTGTG-3′
	Reverse: 5′-GGGGAAGTAGTGATCGTAGCAC-3′
Human_GJB3	Forward: 5′-GATCCAGACCCAGGCAATAACAAGC-3′
	Reverse: 5′-CACTCAGCCCCTGTAGGACCTCTC-3′
Human_GJB4	Forward: 5′-TCAATCGCACCAGCATTAAG-3′
	Reverse: 5′-GGGGGACCTGTTGATCTTATC-3′
Human_GJB5	Forward: 5′-GCTGCTTGCTGAGTCCTATTGCC-3′
	Reverse: 5′-TCCACGCTCGCCTTGAACACTAG-3′
Human_GAPDH	Forward: 5′-TGACTTCAACAGCGACACCCA-3′
	Reverse: 5′- CACCCTGTTGCTGTAGCCAAA-3′

The analysis of GJB2, GJB3, GJB4, and GJB5 at the single-cell level

We utilized CancerSEA to explore the functional states of GJB2, GJB3, GJB4, and GJB5 in various cancer types, and displayed them in heatmaps. This enabled us to analyze their correlation with multiple functional states of cancer cells at the single-cell level. The results indicate that GJB2 is negatively correlated with most biological processes, such as DNA damage or repair, EMT, hypoxia, invasion and metastasis, in GBM and UM tumors. In LUAD, it is positively correlated with DNA damage or repair, EMT, hypoxia, metastasis and proliferation, but without statistical significance (Fig. 7A). The results indicate that GJB3 is positively correlated with cell cycle, DNA damage, EMT, hypoxia, invasion and metastasis in NSCLC, and negatively correlated with apoptosis and quiescence (Fig. 7B). The results demonstrate that both GJB4 and GJB5 are mainly enriched in GBM and HNSCC, and they exhibit similar functions, being negatively correlated with DNA damage or repair, EMT, invasion and metastasis. In HNSCC, they are negatively associated with cell cycle and DNA damage or repair, while positively correlated with metastasis and stemness (Fig. 7C and D). The UMAP plot of the LUAD UNB 10X GSE131907 dataset from the IMMUcan database is shown in Fig. 6E. The expression of GJB2, GJB3, GJB4, and GJB5 were visualized on a UMAP plot, revealing their predominant distribution in the malignant regions (Fig. 7F-I).

Fig. 7.

The expression levels of GJB2, GJB3, GJB4 and GJB5 at single-cell levels A–D: The relationship between GJB2, GJB3, GJB4 and GJB5 expression and different functional states in tumors was explored by the CancerSEA tool. *p < 0.05; **p < 0.01; ***p < 0.001. E: UMAP plot of LUAD_UNB_10X_GSE131907 dataset from IMMUcan. The cells are colored according to their major annotation; F-I: Expression of GJB2, GJB3, GJB4 and GJB5 visualized on a UMAP plot from IMMUcan

More sensitivity to chemotherapies for high expression of GJB2, GJB3, GJB4 or GJB5 subtype

Using the pRRophetic algorithm and the "pRRophetic" drug package, we conducted an analysis of the classic LUAD chemotherapeutic agents, paclitaxel and cisplatin, to predict the IC50 of each sample in the high and low expression groups of GJB2, GJB3, GJB4, and GJB5, respectively (Fig. 8). The results revealed that the IC50 values for paclitaxel and cisplatin were lower in the high expression groups of GJB2, GJB3, GJB4, and GJB5, indicating that patients with high expression of GJB2, GJB3, GJB4, or GJB5 are highly sensitive to these two classic LUAD chemotherapeutic agents.

Fig. 8.

The distribution of IC50 score The abscissa represents different groups of samples, G1 represents the high expression group of GJB2, GJB3, GJB4, and GJB5, while G2 represents their corresponding low expression group and the ordinate represents the distribution of the IC50 score. Different colors represent different groups, top-left represents the significance p-value test method. A–D represents the paclitaxel group and E, F represents the cisplatin group. *p < 0.05, **p < 0.01, ***p < 0.001, asterisks (*) stand for significance levels. The statistical difference of two groups was compared through the Wilcox test, significance difference of three groups was tested with Kruskal–Wallis test

Discussion

The intricacy of Cx biology has established a foundation for investigating the contribution of Cx and gap junction intercellular communication (GJIC) in the pathogenesis of several diseases, including cancer. However, to our knowledge, no prior studies have explored the involvement of genes from the gap junction protein family in cancer progression genetic programming, despite the crucial roles played by some members of the Cx family in various types of cancer. In this work, we examined the predictive value and differential expression of various Cx family members in pan cancer. We found that different tumor types display different expression and survival patterns for members of the connexin family. Prognostic analysis of Cx family members and expression differences holds greater significance in LUAD. To develop a prognostic model that integrates clinical data and mutations associated with prognosis in the Cx family, we investigated the expression, mutation, and prognostic value of various Cx family members in LUAD. In this regard, we established a risk model for lung adenocarcinoma based on the expression of GJB2, GJB3, GJB4, and GJB5.

Although ample evidence supports the role of connexins as tumor suppressors, a number of exceptions to this theory have arisen. By integrating expression data with prognostic analysis, our study revealed that connexins can function as both tumor suppressor genes and oncogenes. The extensive reduction of gap junction intercellular communication (GJIC) in solid tumors provides evidence for the negative correlation between connexin expression and tumor progression (Aasen et al. 2016). Furthermore, our research indicates that in certain cases, connexin expression may promote tumor aggressiveness, which is in line with Ansen‘s study (Aasen et al. 2019). The expression of Cx family members in different cancers displays variations as reported by the TCGA database, with high expression of GJA3 in LUSC and low expression of GJA3 in KIRC. Interestingly, high expression of connexins not only exhibits an anti-tumor effect with improved prognosis, but also promotes tumor growth with poor prognosis, as indicated by our findings. Notably, GJB2, GJB3, GJB4, and GJB6 were significantly upregulated in lung adenocarcinoma, whereas GJA1, GJA4, GJA5, GJC1, GJC2, and GJD3 showed lower expression. In the LUAD transcriptome, GJB2, GJB3, GJB4, GJB5, and GJD3 showed strong associations with other members of the family.

Studies have shown that the higher mRNA and protein levels of connexins observed in various tumors are often associated with connexin mislocalization (Ezumi et al. 2008; Kyo et al. 2008). Cells typically express multiple connexin isotypes simultaneously. Hence, evaluating the expression and correlation of various Cx family members provides a basis for selectively modulating the expression of different connexin isotypes. The differences in Cx expression in LUAD can influence intercellular communication. Gap junctions facilitate intercellular communication by allowing direct exchange of small cytoplasmic molecules between adjacent cells. This non-selective exchange is controlled by gap junction intercellular communication (GJIC) and is involved in a variety of physiological processes. The mechanism of action of cell gap junction proteins is closely related to this phenomenon. Our study found that GJB2, GJA5, and GJA1 are involved in promoting EMT, while GJB1 plays a role in suppressing apoptosis. Additionally, GJC2, GJB1, and GJA5 were found to inhibit the cell cycle, and GJB1 was shown to prevent EMT. Due to the ability of cells to co-express and mix different connexin isotypes, there are several possible combinations of connexin activities. Furthermore, our research demonstrated that GJB3 and GJB2 inhibited both the TSC-mTOR and PI3K-AKT pathways. The differential pathway activity between the high and low GSVA-score groups of differentially expressed connexins was responsible for the inhibition of the TSC-mTOR and PI3K-AKT pathways. While pannexins and connexins do not share sequence homology, they possess comparable structures, including four hydrophobic transmembrane domains, three cytoplasmic domains such as the N- and C-terminal tails, and two extracellular loops. The pannexin family encompasses Pannexin 1 (PANX1), Pannexin 2 (PANX2), and Pannexin 3 (PANX3) as members (Bruzzone et al. 2003). After conducting a protein–protein interaction (PPI) network analysis on the Cx family, we found a robust correlation between the Cx family and the Pannexins family. This suggests the possibility of a close functional relationship between these two protein families that may contribute to the development of cancer.

We conducted an extensive analysis of the Cx family including mutation frequency, CNV changes, and DNA methylation using data from the TCGA and GSCA databases. Our findings indicate that dysregulation of Cx expression is commonly observed in LUAD, with mutations in GJA8, GJA10, GJA9, GJB4, and GJA1 being more frequent. Additionally, we found a positive association between Cx mRNA expression and DNA methylation. To further investigate the involvement of Cx family members in the prognosis of LUAD, we performed Univariate and multivariate Cox regression analyses and developed a diagnostic risk model based on GJB2, GJB3, GJB4, and GJB5 genes. Our results demonstrate that the mutation rates of these four genes are higher in the high-risk population and strongly correlated with differences in OS in LUAD patients. Previous studies have demonstrated that members of the Cx family can serve as potential prognostic indicators and therapeutic targets for cancers with poor survival rates. Moreover, they have also been linked to chemotherapy resistance and the immunological microenvironment. For example, it has been demonstrated that cytoplasmic GJB2 is associated with lymph node metastases and poor prognosis (Naoi et al. 2007). The connexin GJB2 has been shown to play a role in the progression of dual carcinoma in situ (DCIS) to invasive ductal carcinoma (IDC) (Liu et al. 2019). Similar to this, GJB2 overexpression increased tumor development, EMT (reduced E-cadherin; elevated Vimentin and Slug), migration, and invasion partly through the PI3K/Akt pathway (Yang et al. 2015), which is consistent with our research results. Recent studies have revealed that GJB3 may stimulate the polarization and survival of neutrophils, resulting in an accelerated hepatic metastasis of pancreatic cancer (Huo et al. 2022). According to Yuanlin et al., elevated expression of GJB3 has been linked to unfavorable prognostic outcomes (Wu et al. 2022). The role of GJB4 in promoting tumor progression, stem cell development, and cancer cell metastasis has been firmly established (Lin et al. 2019). GJB4 can also enhance migration and proliferation of stomach cancer cells by activating the Wnt/CTNNB1 pathway. Overexpression of GJB5 in H1299 cells impeded migration and invasion, while GJB5 knockdown restored these cellular processes (Zhang et al. 2012). According to the findings of our research, this differs from the function of GJB5, possibly due to varying behavior of different cell lines.

We developed a risk model based on the four Cx features GJB2, GJB3, GJB4, and GJB5, which reliably predicts prognosis in patients with LUAD. Our model maintained its reliability when tested against the GEO dataset GSE31210. Furthermore, using RT-qPCR, we assessed the expression of the four Cx proteins in four lung adenocarcinoma cell lines (A549, PC-9, HCC827, and NCI-1975) in comparison to the healthy human bronchial epithelial cell line Beas2B. All of the four Cx showed high expression in lung adenocarcinoma cell lines. Although high expression of GJB2, GJB3, GJB4, and GJB5 is associated with poor prognosis, the results of chemotherapeutic response prediction suggest that the high expression of these subtypes is more sensitive to classic LUAD chemotherapeutic agents, such as paclitaxel and cisplatin. Therefore, distinguishing the high expression of GJB2, GJB3, GJB4, or GJB5 subtypes in the population of lung adenocarcinoma patients not only predicts prognosis but also provides a reference for the treatment of these patients. We hypothesize that the sensitivity of Cx family members to chemotherapeutic agents may not be sufficient to counteract their contribution to various malignant behaviors of tumors, including EMT, ECM and surface adhesion, cellular migration and motility, angiogenesis, and adherens junctions. Therefore, exploring the enhancement of multiple malignant behaviors of tumors by Cx is essential. Further investigation into the role of Cx family members and the combined use of corresponding drugs is of significant clinical importance and may contribute to the development of more effective treatment strategies.

We conducted further analysis on the expression and biological functions of GJB2, GJB3, GJB4, and GJB5 at the single cell level. In LUAD, GJB2 displays a positive correlation with DNA damage or repair, EMT, hypoxia, metastasis, and proliferation. GJB3 is positively correlated with the cell cycle, DNA damage, EMT, hypoxia, invasion, and metastasis of NSCLC. On the other hand, it is negatively correlated with cell apoptosis and quiescence. Although GJB4 and GJB5 are not enriched in the biological behavior related to lung cancer, they exhibit similar functions and are negatively correlated with DNA damage or repair, EMT, invasion, and metastasis. Recent research suggests that Cx31 displays adhesion-independent activation of the FAK signaling pathway, inducing FAK phosphorylation in a non-adherent manner. Moreover, Cx31 is involved in the activation of the NF-kB signaling pathway, which promotes the survival of metastatic cells in the brain, in a FAK-dependent manner (Lorusso et al. 2022). Based on the UMAP graph, it appears that GJB2, GJB3, GJB4, and GJB5 are primarily expressed in the minor regions. Furthermore, since Cx family members are often associated with EMT, invasion, and metastasis, we utilized the pan-cancer dataset to compute the correlation matrix between each Cx family member and genes related to EMT, cell adhesion and ECM, migration and motion, as well as angiogenesis. Our findings suggest that not only are most Cx members associated with the EMT genetic program, but they are also connected to the transcriptional program that promotes ECM production, migration, movement, and angiogenesis. The invasion-metastasis cascade refers to a complex, multi-step mechanism that leads to the development of metastasis in cancer patients (Valastyan and Weinberg 2011). Metastasis is a complex process involving several steps that cancer cells must undergo in order to leave their primary site and establish secondary tumors in other organs. These steps include the ability of cancer cells to break through the basement membrane and move through the lymphatic or blood vessels. Once in circulation, cancer cells must survive and attach to the endothelium, and then transmigrate into the target organ. To establish secondary tumors, disseminated tumor cells must endure and proliferate in the target organ (Riggi et al. 2018).

Conclusions

The heterogeneity in expression patterns of Cx family members across various tumors and their prognostic implications are intricate. While the assessment of Cx expression in human tumors is feasible as a prognostic marker, due attention should be paid to specific analyses for different cancer types, and a precise patient-tailored strategy should be employed to avoid unnecessary adverse effects. By scrutinizing the functional role of Cx in tumorigenesis, it has been discovered that the expression levels of GJB2, GJB3, GJB4, and GJB5, in conjunction with clinical data, can serve as a prognostic indicator for OS in patients. Additionally, these Cx may facilitate genetic programs involved in the transcriptional regulation of EMT, ECM production, migration, motility, and angiogenesis by influencing the PI3K-AKT and TSC-mTOR signaling pathways. Thus, the levels of GJB2, GJB3, GJB4, and GJB5 expression hold potential as prognostic markers and therapeutic targets for the evaluation of poor survival in human lung adenocarcinoma tumors. Nevertheless, the functional role of GJB3, GJB4, and GJB5 in lung adenocarcinoma remains obscure, and further investigation into their potential as novel biomarkers and therapeutic targets could play a crucial role in improving patient outcomes and reducing tumor metastasis.

Enhanced understanding of the complexity of cancer biology related to connexin and cell–cell communication may lead to the design of novel therapeutic strategies. Regulation of connexin expression could be an effective therapeutic approach for lung adenocarcinoma, with the potential to improve immune response and drug resistance in affected patients.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

PJ and NZ: conceived the study. Material preparation, data collection and analysis were performed by PJ, XH and BD. The first draft of the manuscript was written by PJ and all authors commented on previous versions of the manuscript. XZ and NZ: completed model guidance, critical review, and funding support. All authors read and approved the final manuscript.

Funding

This work was supported by a grant obtained from the Qilu leader training project (Na Zhou).

Data availability

All data generated or analysed during this study are included in this published article and its supplementary information files.

Declarations

Conflict of interest

The authors have no relevant financial or non-financial interests to disclose.

Ethical approval

Not applicable.

Consent to participate

Not applicable.