Systematic expression analysis of the CELSR family reveals the importance of CELSR3 in human lung adenocarcinoma

Abstract Cadherin EGF LAG seven‐pass G‐type receptors (CELSRs) are involved in the progression of various types of cancer. CELSR3, a crucial signalling molecule in the WNT/PCP pathway, is believed to be associated with tumorigenesis and metastasis. However, its role in lung adenocarcinoma (LUAD) remains unclear. In this paper, we analysed the expression of CELSR family members using the Oncomine, GEPIA and UALCAN databases. We used a Kaplan‐Meier plotter to assess the effect of CELSRs on tumour prognosis. Next, gene ontology (GO), KEGG pathway, miRNA target, kinase target and transcription factor‐target enrichment were analysed by GSEA. Simultaneously, we conducted functional assays including cell viability, colony formation and transwell assays, to determine the oncogenic role of CELSR3 in LUAD. Finally, we used the TIMER and TISIDB databases to analyse the correlation between CELSR3 and immune infiltration and the potential chemokine receptor axis causing immune cell expression. High expression of CELSR3 is in LUAD predicts poor prognosis and early progression of the tumour. KEGG and GO enrichment analysis revealed the functional relationship between CELSR3 and cell adhesion, the cell cycle, and DNA replication. Down‐regulation of CELSR3 suppressed cell proliferation to a significant extent, in addition to inhibiting invasion and migration in LUAD cells. Finally, CELSR3 expression was significantly correlated with the infiltration level of CD8+T cells through the CCL17/CCR4 axis in LUAD. These results indicate that CELSR3 can serve as a prognostic biomarker for determining prognosis and immune infiltration in LUAD.

be involved in the regulation of contact-dependent neurite growth 9 and may play an important role in the development and metastasis of different cancers. 10 The Wnt signalling pathway is thought to play a key role in tumorigenesis through the typical Wnt/catenin cascade.
However, a growing body of evidence has revealed the previously unrecognized role of non-canonical Wnt/planar cell polarity (PCP) 11 signalling in cancer progression, metastasis, evasion and angiogenesis.
Thus, CELSR3, as a crucial signalling molecule in the WNT/PCP pathway, is believed to be associated with tumorigenesis and metastasis.
However, the role of CELSR3 in the LUAD development remains unknown, as does whether CELSR3 is a specific marker in LUAD. In this study, we used the Oncomine, GEPIA and UALCAN public databases to analyse the CELSR family expression. Then, we used the Kaplan-Meier(KM) Plotter to evaluate the association between CELSR family mRNA levels and disease prognosis in patients with LUAD. In addition, we explored the relationship between CELSR3 mRNA expression levels and clinical subgroups and tumourinfiltrating immune cells in patients with LUAD.

| Oncomine database analysis
Oncomine database is an online cancer microarray database that aimed at facilitating discovery from genome-wide expression analyses (https://www.oncom ine.org/resou rce/login.html). 12,13 The expression levels of CELSR mRNA (log2-transformed) in lung cancer tissues and normal tissues were searched using Oncomine database, and the transcription levels of CELSRs in all lung cancer tissues were determined for statistical comparison. To obtain the important CELSR probes, we set the following thresholds: P-value < .05, fold change >1.5 and gene ranks in the top 10%.

| UALCAN analysis
UALCAN is an interactive web resource that provides comprehensive cancer transcriptome date for in-depth analysis of The Cancer Genome Atlas (TCGA) gene expression and clinical data from 31 types of cancer (http://ualcan.path.uab.edu ). 14,15 Using UALCAN, we also analysed the relative CELSR3 expression in LUAD across tumours and corresponding normal tissues, in addition to different tumour subgroups.

| GEPIA analysis
The GEPIA (http://gepia.cance r-pku.cn/index.html) 16 is an interactive website based on TCGA and GTEx-based project, where RNA sequencing expression in tumours and corresponding normal tissues can be analysed. We used GEPIA to analyse the correlation of TCGA expression data and the Spearman method to determine the correlation coefficient between two genes.

| KM Plotter survival analysis
The prognostic value of the CELSRs mRNA expression was assessed by the Kaplan-Meier plotter (www.kmplot.com), 17 an online database containing gene expression data and survival information for lung cancer patients. Using this database, we analysed the overall survival (OS) and first progression (FP) of patients with LUAD. The patient samples were divided at the median expression into two groups with either high expression or low expression, and evaluated using the Kaplan-Meier survival plot [P-value < .05, false detection rate (FDR) < 0.05).
The risk ratio (HR) had a 95% confidence interval and log-rank P-value.

| LinkedOmics analysis
The LinkedOmics database is a website used to analyse 32 TCGA cancer-related datasets (http://www.linke domics.org/login.php). 18 We used the LinkFinder module of LinkedOmics to study the differentially expressed genes related to CELSR3 in the TCGA LUAD (n = 515). Pearson correlation coefficient was used for statistical analysis of the results, and LinkFinder was used to establish a statistical diagram of a single gene. Functional enrichment analysis was applied using the Web-based Gene SeT Analysis Toolkit (WebGestalt). 19 The data in the LinkFinder results were signed and sequenced, and we used gene set enrichment analysis (GSEA) for gene ontology (GO) analysis, KEGG and network analysis. We used the Molecular Signatures Database (MSigDB) 20 to analyse the network (FDR <0.05; and 500 simulations were performed).

| Tumour immune estimation resource (TIMER) database analysis
Tumour immune estimation resource is a comprehensive resource for the systematic analysis of immune infiltrates across 32 tumour types (https://cistr ome.shiny apps.io/timer/). 21 TIMER employs a previously published statistical method 22 to infer the abundance of tumour-infiltrating immune cells (TIICs) from gene expression profiles. We analysed the expression of CELSR3 in different types of lung cancer and the correlation of CELSR3 expression with infiltration of immune cell, including B cells, CD8+ T cells, CD4+ T cells, macrophages, neutrophils and dendritic cells (DCs), through gene patterns. 23 Moreover, we studied the correlations between CELSR3 expression and genetic markers in TIICs through related modules.

| TISIDB analysis
The TISIDB is a web portal (http://cis.hku.hk/TISIDB) that combines 4176 records from 2530 publications and reports on 988 genes associated with anti-tumour immunity. We can explore the function of genes of interest and their role in tumour immune interactions by analysing the TISIDB database and through high-throughput data analysis and literature mining. 24 On the X-axis, CELSR3 is used as the gene symbol, and on the Y-axis, the TIICs-related gene marker is used as the gene symbol. We used Log2 RSEM to determine the gene expression level.

| Cell culture and Small Interfering RNA Transfection
We obtained the LUAD A549 and H1975 cell lines from the Cell Bank of Shanghai Biology Institute (Shanghai, China). We cultured these cells in RPMI-1640 medium supplemented with 10% calf serum (Invitrogen, USA), 100 IU/mL penicillin and 100 IU/mL streptomycin at 37℃ in a humidified atmosphere with 5% CO 2 . For transfection, we cultured the cells to 70% confluence and transfected them with siRNA using  Table S4. The si-CELSR3#1 sequences were used in cell transfection. The primers used in qRT-PCR are listed in Table S5.

| Cell counting kit-8 assay
We transfected A549 and H1975 cells were with either CELSR3 siRNA or an siRNA control (GenePharma, Shanghai, China).
Following transfection for 24 hours, the cells were seeded into a 96well plate at 5.0 × 10 3 cells/mL and continuously cultured for 24, 48, and 72 hours. At each time-point, we added CCK8 reagent (DojinDO, Shanghai, China) to each well and incubated for 2 hours at 37°C. We measured spectrometric absorbance at a wavelength of 450 nm on a microplate reader (Spectra Max M5, MD, USA). Each sample was tested in triplicate, and all experiments were performed three times.

| Colony formation assays
We transfected A549 and H1975 cells with either CELSR3 siRNA or an siRNA control (GenePharma, Shanghai, China). After 24 hours transfection, we seeded 2.5 × 10 3 A549 cells and 5 × 10 3 H1975 cells into a six-well plate. To achieve colony formation, we grew the cells for 6 days in normal culture medium. We stained the colonies with crystal violet.

| Transwell assay
We used a Boyden chamber assay to study cell invasion capability. Cells were transfected with either CELSR3 siRNA or an siRNA control (Ruibobio, Shanghai, China). After 24 hours, we trypsinized and resuspended the transfected cells and placed 5.0 × 10 4 cells in 200 μL RPMI-1640 medium into the upper chambers (8-mm pore size; Millipore). We filed the lower chambers with 600 μL complete medium containing 10% FBS. Following 24-hour incubation of the cells at 37°C, we used a cotton swab to remove non-invading cells from the top of the chamber. The invading cells on the lower surface of the inserts were fixed and stained with 0.1% crystal violet, and five random fields for each insert were counted at 100× magnification.

| Statistical analysis
All statistical analyses were performed using online resources.
Survival curves were generated using the Kaplan-Meier method. The results generated in Oncomine are displayed with Student's t test to compare mRNA expression. Kaplan-Meier plots are shown with a hazard ratio with 95% confidence intervals and log-rank P-value. GEPIA variance analysis was tested using one-way ANOVA by defining the disease status (tumour or normal) as variable. All data are presented as the mean ± standard derivation (SD). SPSS 19.0 software (SPSS, Chicago, IL, USA) was applied for the statistical analyses. Spearman's correlation was used to evaluate the correlation of gene expression, and all P-values < .05 were considered statistically significant.

| Up-regulation of CELSR3 mRNA expression in human LUAD
Three CELSR family members have been identified in human cancers. We analysed the transcriptional levels of the CELSRs using the Oncomine database in different tumours and normal tissues of multiple cancer types. Our findings showed that the expressions of CELSR3 and CELSR2 were higher in lung tumours tissues compared with the corresponding normal tissues ( Figure 1A). Further Oncomine analysis showed that many studies suggested high CELSR3 expression in LUAD (Table 1). Then, we mined the GEPIA and UALCAN databases, and found that CELSR3 was the only higher CELSR family member relative to normal tissues in LUAD tissues ( Figure 1B Moreover, in subgroup analyses based on sex, age, ethnicity, stage and tumour grade, the CELSR3 transcription was significantly higher in LUAD patients than in healthy individuals ( Figure 2).

| High CELSR3 expression predicts poor prognosis in LUAD
We used the KM plotter to analyse the correlation between the mRNA

| GO and KEGG pathway analyses of coexpression genes correlated with CELSR3 in LUAD
We analysed the mRNA sequencing data of 515 patients with LUAD in TCGA were analysed using the LinkedOmics functional module.
The volcanic diagram ( Figure 4A) showed that 2832 genes (dark red dots) were significantly positively correlated with CELSR3, and 2275 genes (dark green dots) were significantly negatively correlated with

| CELSR3 networks of kinase, miRNA and transcription factor targets in LUAD
Gene expression is tightly regulated by multiple layers of mechanisms, including kinase, miRNA and transcription factor (TF) binding. Using GSEA, we studied the kinases, miRNAs and transcription factor-target networks of CELSR3 related gene sets to explore the mechanism of action of CELSR3. In general, the seven most significant target networks were kinase-target networks related mainly and in a subtle fashion to the kinases ATR, ATM/PLK1, CDK1, CDK2, CHEK1 and CHEK2 (

| Silencing of CELSR3 in LUAD cells inhibits cell proliferation and motility
To explore the relationship between CELSR3 and proliferation, cell  Figure S1). Hence, in the subsequent analysis, we used si-CELSR3#1 for specific RNA interference ( Figure S4).

| CELSR3 knockdown inhibits proliferation and ablation of CELSR3 suppresses migration and invasion in LUAD cell
We used a CCK8 assay to study how CELSR3 expression influences LUAD cell proliferation. We performed knockdown of CELSR3 in the lung cancer cell lines A549 and H1975. The results suggested that CELSR3 significantly affected cell proliferation ( Figure 5A-B). As shown in Figure 5A, the cell number increased in all groups over time, and group was visibly less than that in NC group ( Figure 5E-F, P < .01).

| Correlation of CELSR3 expression with infiltrating CD8+ T cells
We analysed the correlation between CELSR3 expression and six

| Correlation of CELSR3 expression with the CCL17/CCR4 axis
We examined the correlations between CELSR3 expression and the

| Correlation of markers of CD8+ T cells with the CCL17/CCR4 axis
To analyse the relationship between CD8+ T cells and the CCL17/ CCR4 axis, we examined the correlations between markers of CD8+ T cells and the CCL17/CCR4 axis in the TIMER database.

| D ISCUSS I ON
Lung adenocarcinoma severely affects human health and is associated with significant morbidity and mortality. Recently, a relationship between the expression of the CELSR family and tumours, such as adult brain tumours 25 and ovarian cancer, 26 actually has also been reported. Although some members of the CELSR family have been shown to play a key role in cancer, the exact roles of CELSR3 in LUAD remained unclear. In this study, we analysed the expression and value of CELSR3 in LUAD. In addition, we also investigated the association between CELSR3 expression and immune infiltration, as well as the molecular mechanism leading to different levels of immune infiltration.
Our findings have shown CELSR3 expression to be significantly higher in LUAD tissues than in adjacent normal tissues. To the best of our knowledge, this is the first report of a consistent relationship between elevated CELSR3 mRNA levels and poor prognosis and early tumour progression in patients with LUAD; although other studies have described the role of CELSR3 in the occurrence and development of several tumour types, including liver 27 and oral squamous carcinomas. 10 These findings indicate that CELSR3 may play an oncogenic role and be an important prognostic indicator in LUAD.
To further explore the functions and mechanism of CELSR3 in LUAD, we finally performed GSEA using the target gene sets to help identify the networks of target kinases, miRNAs and transcription factors. CELSR3, as the key crucial signalling molecule in the WNT/ PCP pathway, is thought to be involved in tumorigenesis and metastasis. 28 The main conclusion we draw from our findings is that the As far as we know, the proliferation and activation of CD8+T lymphocytes directly kill tumour cells and have anti-tumour effect. 42 Indeed, successful delivery of CD8+ effector T cells across the tumour vessels is considered to be a key determinant of anti-tumour immunity. 43 Although CD8+ cytotoxic T lymphocytes (CTLs) play an important role in tumour regression, CD4+ cells are necessary for CTLs to enter tumour tissue. 44 Furthermore, our results indicated a certain direction for CELSR3 in the level of immune cell infiltration.
In summary, our results showed that CELSR3 plays an important role in regulating CD8+T cells infiltration in LUAD.
Numerous studies have shown that the interaction between chemokines and cell receptors allows targeted function of circulating immune cells. Different types of immune cell infiltration are closely regulated by chemokines that regulate tumour immunity and biological phenotypes of tumours, in addition to influencing tumour progression, therapy and prognosis. 45 To elucidate effector T cell differentiation and protective immunity, we studied the chemokines and receptors associated with CELSR3 in the TISIDB database and performed further analysis using the GEPIA, KM plotter, and TIMER databases (Table S1-S3). Interestingly, we found that CCL17/ CCR4 was the only axis associated with CELSR3 expression that was statistically significant. Using the TISIDB database, we studied relationships between the CCL17/CCR4 axis and CELSR3 expression, copy number, methylation and mutation. Our findings revealed negative correlation between levels of expression of the CCL17/CCR4 axis and CELSR3 expression; this trend was reversed when CELSR3 was either methylated or mutated. This is the first time, as far as we know, that CELSR3 influences the infiltration of CD8+T cells through the CCL17/CCR4 axis. The CCL17/CCR4 axis is known to basically regulate gene expression and play an important role in T cell differentiation and function, which is fairly significant. Previous studies have shown that the CCL17/CCR4 receptor axis represents a group of potential new targets for the treatment and prevention of autoimmune diseases of the central nervous system. 46 Indeed, DCderived CCL17 was found to promote the interaction between DCs and CD8+ T cells, thus activating CD8+ T cells. 47,48 We have found that CD8+ T cells to be significantly and positively associated with the CCL17/CCR axis (P < .05). These findings are in accordance with those of our previous reports. Moreover, the CCL17/CCR4 axis may play an essential role in the differentiation and function of CD8+ T

cells.
This study has several important limitations. First, as it has been based on data retrieved from published articles and public repositories, the quality of the data can influence the study outcomes. In in vitro experiments, we only verified the function of proliferation, invasion and metastasis. We will further verify the results of the immunocorrelation analysis in in vivo experiments. Second, the number of samples in the database is continuously monitored and expanded, which may affect the results of our study. Consequently, we aim to collect more clinical cases to verify the prognostic value of CELSR3 in LUAD. Third, the accuracy and selection of the statistical methods used by the database to analyse the data may influence the interpretation of the research results. More effective analytical means and methods need to be learned and developed. However, we obtained similar results from analysing multiple databases, thus supporting our research conclusions. In future analyses, additional basic and clinical trials will be necessary to verify the mechanism of action and prognostic value of CELSR3 in LUAD.

| CON CLUS IONS
To the best of our knowledge, this is the first study to demonstrate high specific expression of CELSR3 mRNA and its prognostic effects in LUAD. Down-regulation of CELSR3 significantly suppressed cell proliferation and dramatically reduced the ability of of LUAD cells to migrate and invade. More importantly, this is the first study to show that CELSR3 may influence the level of infiltration level of CD8+ T cells through the CCL17/CCR4 axis in LUAD. Obviously, in future analyses, more basic and clinical trials will be required to verify diagnostic and immunotherapeutic value of CELSR3 in LUAD.

ACK N OWLED G EM ENTS
This study was supported by General Project of Natural Science