A comprehensive analysis of FBN2 in bladder cancer: A risk factor and the tumour microenvironment influencer

Abstract Bladder cancer (BLCA) is a common and difficult‐to‐manage disease worldwide. Most common type of BLCA is urothelial carcinoma (UC). Fibrillin 2 (FBN2) was first discovered while studying Marfan syndrome, and its encoded products are associated with elastin fibres. To date, the role of FBN2 in BLCA remains unclear. The authors first downloaded data from The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO). The patients were divided into high FBN2 expression and low FBN2 expression groups, and the survival curve, clinical characteristics, tumour microenvironment (TME), and immune cell differences were analysed between the two groups. Then, the differentially expressed genes (DEGs) were filtered, and functional enrichment for DEGs was performed. Finally, chemotherapy drug susceptibility analysis based on the high and low FBN2 groups was conducted. The authors found upregulated expression of FBN2 in BLCA and proved that FBN2 could be an independent prognostic factor for BLCA. TME analysis showed that the expression of FBN2 affects several aspects of the TME. The upregulated expression of FBN2 was associated with a high stromal score, which may lead to immunosuppression and be detrimental to immunotherapy. In addition, the authors found that NK cells resting, macrophage M0 infiltration, and other phenomena of immune cell infiltration appeared in the high expression group of FBN2. The high expression of FBN2 was related to the high sensitivity of some chemotherapy drugs. The authors systematically investigated the effects and mechanisms of FBN2 on BLCA and provided a new understanding of the role of FBN2 as a risk factor and TME influencer in BLCA.

as a highly effective form of intravesical treatment for NMIBC [5]. However, some of the patients treated with BCG may experience non-response, recurrence, severe adverse reactions, or remain progress to muscle invasive bladder cancer (MIBC) [6,7]. Radical cystectomy, even chemotherapy, immune checkpoint blockade (ICB) therapy is often required for MIBC patients. However, there are still some adverse effects of ICB therapy [8][9][10].
Tumour cells, stromal cells, and tumour-infiltrating immune cells make up the majority of the TME [11]. According to research, stromal components can form the TME, impact chemotherapy and immunotherapy response, and encourage cancer progression [12]. To solve the bottleneck of immunotherapy, it is urgent to study the TME of BLCA, which is also the focus of current research.
FBN2 is a 279.57 kb protein-coding gene mapping in 5q23-31. The protein that FBN2 encodes is associated with elastin fibres, which are associated with the matrix [13]. Past studies on FBN2 mainly focused on congenital contractural arachnodactyly (CCA), and it has been proven that mutation in FBN2 is an important cause of CCA [14]. Recently, some studies have found a possible link between FBN2 and some types of tumours, including colorectal adenocarcinomas and non-small-cell lung cancer [15,16]. High expression of FBN2 may be associated with TGF-β signal enhancement [17]. TGFβ is generally considered to be associated with tumour progression and immunosuppression in the TME [18,19]. This suggests that FBN2 may be correlated with the TME. It has been found that the epithelial-mesenchymal transition or TME related gene prognostic models containing FBN2 have good prognostic performance [20,21]. This implies that FBN2 plays an important role in BLCA. However, the effect of high or low expression of FBN2 on TME and the efficacy of chemotherapy drugs in BLCA remains unclear. We conducted an indepth study on the prognostic properties of FBN2 in BLCA and its influence on immune and chemotherapy drugs in order to gain a comprehensive understanding of the role of FBN2 in bladder cancer.

| Analysis of FBN2 differential expression
TIMER [22] (http://timer.comp-genomics.org/) is a powerful web analysis tool that can perform pancancer analysis for the differential expression levels with Wilcoxon testing of particular genes between tumours and adjacent normal tissues based on TCGA, and then box plots are used to visualise the results [22]. To systematically understand the expression levels of FBN2 in different tumours, TIMER was used to analyse the expression levels of FBN2 in different tumours that TCGA contained.
For the expression of FBN2 in BLCA, the R package 'limma' [23] was used to analyse and visualise the difference in FBN2 expression between BLCA and normal tissues. The correlation between FBN2 and other genes was explored with the filter criteria set to a correlation coefficient >0.4 and p value <0.001. The R package 'circlize' [24] was used to draw the association networks of FBN2 and other genes.

| Survival analysis based on FBN2 expression
Patients were divided into a high FBN2 expression level group and a low FBN2 expression level group on the basis of the median FBN2 expression. Kaplan-Meier (KM) analysis was performed for TCGA samples and GEO samples with the end events set to death, and log-rank tests were performed to determine the p value. KM analysis was completed by the R package 'survival' (https://CRAN.R-project.org/package= survival).
The receiver operating characteristic (ROC) curve was used to evaluate the accuracy of FBN2 for predicting prognosis. The area under the curve (AUC) was an effective method to summarise the accuracy of the ROC, and the higher the value of the AUC (no more than 1), the better the predicting ability. The AUCs at 1, 3, and 5 years were calculated. ROC analyses were completed by the R package 'timeROC' [25].

| Correlation analysis between FBN2 and clinical features
To analyse the relationship between FBN2 expression levels (high FBN2 expression and low FBN2 expression) and clinical features in detail, the R package 'limma' [23] was used. The clinical features included in the analysis were age, gender grade, stage, and TNM stage. The R packages 'ComplexHeatmap' [26] and 'ggpubr' (https://CRAN.R-project.org/package=ggpubr) were used to visualise the results. To further test whether the FBN2 expression levels and clinical features were independent prognostic factors, univariate and multivariate Cox regression were executed, and forest plots were created to present results visually based on the R package 'survival' (https://CRAN.Rproject.org/package=survival). LU ET AL. -163

| Screening of DEGs
To explore the underlying mechanism of the difference between the high FBN2 expression group and the low FBN2 expression group, the DEGs between the two groups should be found for follow-up analysis. The R package 'limma' [23] was used to calculate and filter DEGs. The filter threshold was set to |logFC| >1 and FDR (false discovery rate) <0.05. The heatmap was created by the R package 'pheatmap' (https:// CRAN.R-project.org/package=pheatmap).

| Analysis of the relationship between FBN2 and the TME
The analysis of the difference in the TME between the high FBN2 expression group and the low FBN2 expression group was based on ESTIMATE, which is a scoring method based on expression data [31]. After using the R packages to obtain the scoring documents, the R packages 'limma' [23], 'reshape2' [32], and 'ggpubr' (https://CRAN.R-project.org/package=ggpubr) were used to plot. Then, the differential analysis of immune cells between the high-low FBN2 expression group was performed based on CIBERSORT [33]. The R packages 'e1071' (https://CRAN.Rproject.org/package=e1071), 'preprocessCore' (https://github. com/bmbolstad/preprocessCore), and 'limma' [23] were used, and the filter condition of the p value was set to 0.05. In addition, correlation analysis between the FBN2 expression and immune cells was executed.
Tumour mutation burden (TMB) is considered to predict the therapeutic effect of immune checkpoint blockade (ICB) therapy, TMB has been proven to be a tumour biomarker in some cancers [34]. Based on the R packages 'limma' [23], 'ggplot2' [30], the relationship between TMB and FBN2 expression was analysed by Spearman correlation analysis.

| Chemotherapy drug susceptibility analysis
To evaluate the FBN2 ability in predicting the clinical response of treatment, we calculated the half maximal inhibitory concentration (IC50) of chemotherapeutic agents commonly used for BLCA, including 'cisplatin', 'docetaxel', 'doxorubicin', 'gemcitabine', 'methotrexate', and 'vinblastine'. The differential IC50 of chemotherapeutic agents between low and high FBN2 expression groups was predicted by the 'pRRophetic' package [35]. Based on gene expression microarray data, the pRRophetic package, which was used to determine IC50 with success in previous research [36][37][38][39], was performed for the prediction of clinical chemotherapeutic response by using statistical models from the gene expression and drug sensitivity data from cell lines in the Cancer Genome Project [40] as a training set.
For BLCA, the differential FBN2 expression in all tumour tissues and normal tissues had a statistical significance with a p value less than 0.001, and in 19 tumour tissues and corresponding normal tissues also had a statistical significance with a p value less than 0.01 (Figure 1 b,c). The correction between FBN2 and other genes was also analysed with a screening condition of a p value less than 0.001. The results showed that the expression of 11 genes was related to FBN2 expression ( Figure 1d). FBN2 had a positive correlation with GNA12, MYH10, ARHGAP29, IARS1, CLIC4, and DENMD1A and a negative correlation with HSD17B2, BTBD16, CYP4F35P, CAPN8, and FABP6.

| High FBN2 expression is associated with low overall survival
The results of the KM analysis showed that the overall survival was higher in the low FBN2 expression level group than in the high FBN2 expression level group based on TCGA (p value = 0.011) or GSE19915 (p value = 0.041) (Figure 2a,c). ROC analysis was used to assess the ability to predict prognosis (Figure 2b,d). For TCGA samples, the AUCs at 1, 3, and 5 years were 0.662, 0.614, and 0.572 respectively. For GSE19915 samples, the AUCs at 1, 3, and 5 years were 0.623, 0.688, and 0.637 respectively. The results of KM analysis verified that high expression of FBN2 was associated with survival risk.

| FBN2 is an independent prognostic factor
The results of the relationship between FBN2 expression levels and clinical features are shown in the heatmap (Figure 3b). Except for grade, FBN2 expression was not significantly associated with clinical features. Therefore, the analysis of the relationship between FBN2 expression and the clinical grade was independently performed, and the p value was calculated to be less than 0.001, which may suggest that FBN2 was related to grade (Figure 3a). The results of the univariate and multivariate Cox regression are shown in forest plots (Figure 3c,d). The results suggested that FBN2, age and stage could be considered independent risk factors.

| Functions enrichment of DEGs
A total of 1039 DEGs were finally filtered between BLCA patients with high and low FBN2 expression, and the results  (Table S1). For biological process (BP) molecular function (MF) and cellular component (CC), the DEGs were mainly associated with cornification, receptor ligand activity and synaptic membrane respectively.
KEGG enrichment analysis was also performed, and the results are shown (Figure 4b). The result of the GSEA based on KEGG enrichment suggested that the functions of DEGs in high FBN2 expression were enriched in systemic lupus erythematosus, neuroactive ligand receptor interaction, and olfactory transduction. While the functions of DEGs in low FBN2 expression were enriched in the linoleic acid metabolism and ribosome (Figure 4c).

| Analysis of the correlation between FBN2 and the TME
Based on ESTIMATE, the stromal score, immune score, and ESTIMATE score (the sum of the stromal score and immune score) for the low FBN2 expression group and the high FBN2 expression group were calculated and compared. The high FBN2 expression level group had significantly higher stromal and ESTIMATE scores than the low expression level group (Figure 4d). The active stroma is considered immunosuppressive, and the possible reason is that the active stroma may prevent immune cells from penetrating tumour tissue [41,42]. This may suggest that high FBN2 expression was associated with immunosuppression.
The difference in immune cell infiltration was explored between the high and low FBN2 expression groups (Figure 5a). Among the results that were statistical significance, we found that plasma cells, monocytes, dendritic cells activated, and mast cells resting were higher in the FBN2 low expression group, while NK cells resting and macrophages M0 were higher in the FBN2 high expression group. The correlation between immune cells and FBN2 expression was explored (Figure 5b). The results of correlation analysis with pvalue <0.05 were macrophages M0, M1, NK cells resting which were positively correlated with FBN2 expression, while NK cells activated, plasma cells, dendritic cells activated, mast cells resting, and monocytes which were negatively correlated with FBN2 expression. Besides, the correlation between FBN2 expression and the immune checkpoint genes was performed  (Figure 5c). Except for TNFRSF14, high FBN2 expression was associated with upregulation of immune checkpoints. For the TMB, FBN2 expression was positively but weakly correlated with TMB based on Spearman statistical method with R = 0.14 and P = 0.006 (Figure 5d).

| Susceptibility analysis for chemotherapy drugs commonly used for BLCA
Except for methotrexate, the chemotherapy drugs commonly used for BLCA, including cisplatin, docetaxel, doxorubicin, gemcitabine, and vinblastine, were all found to have statistically significant differences in drug susceptibility between the high and low FBN2 expression groups (Figure 6 a-f). The IC50 values of the 5 chemotherapy drugs were all lower in the high FBN2 expression group than in the low FBN2 expression group. Commonly, a lower IC50 indicates higher drug sensitivity. Therefore, the result may suggest that high FBN2 expression group was more sensitive to these chemotherapy drugs.

| DISCUSSION
In our study, the Pan-cancer analysis for FBN2 was first performed. We found that there were significant differences between some types of tumours and corresponding normal tissues, including BLCA. The data of BLCA patients and normal tissues were downloaded from TCGA and GSE19915 for follow-up analysis. KM analysis for high and low FBN2 expression levels was performed, and the results showed that high FBN2 expression was associated with low overall survival. Moreover, we explored the correlation between FBN2 and clinical features, and the results proved that FBN2 can be a risk factor for BLCA. Thus, we demonstrate that over-expression of FBN2 may be a risk factor for survival in patients with BLCA. The low survival rate in patients with high FBN2 expression may be related to immune escape caused by activation of the immune stroma according to the results of TME analysis. In addition, the results of our susceptibility analysis showed that high FBN2 expression was associated with the better sensitivity of five chemotherapeutic agents in the treatment of BLCA, which may suggest that these chemotherapeutic agents were more suitable for BLCA patients with high FBN2 expression.
Currently, more in-depth research has been done on the involvement of FBN2 primarily in colon cancer. According to clinical studies, 63% of patients with primary colorectal cancer had the FBN2 gene methylated in their tumour samples; this suggests that FBN2 may be a novel diagnostic for the hepatic metastases of colorectal cancer [43]. The FBN2 has a significant potential to detect BLCA, despite the fact that little is known about its biological role in relation to epigenetic alterations in human malignancies. In our study, we discovered that FBN2 and increased FBN2 were consistently associated with shorter BLCA patient survival. This finding stimulates the notion to further explore the function of FBN2 in other domains. Similar to this, elevated FBN2 expression was discovered to be a risk factor for stomach and lung cancer [44,45]. Additionally, it also has been demonstrated that FBN2 might serve as tumour-suppressive effects and is a characteristic basement membrane marker in several types of cancer [46,47]. Additionally, it has been suggested that FBN2 may be a highly effective diagnostic tool for rhabdomyosarcoma and smooth muscle sarcoma [48].
Keratins are encoded by 54 genes, which are the major subgroup of intermediate filament (IF) proteins [49]. IFs are -169 involved in the formation of cytoskeletal systems and play an important role in maintaining the stability of epithelial tissue [50]. We found that the functions of DEGs were mainly focused on cornification in GO enrichment. UC, the major type of BLCA, is a cancer of the epithelial cells destined to be closely related to keratin. Different molecular subtypes of UC commonly have different keratin levels [51], and keratin expression profiles have been recognized as diagnostic and prognostic markers for BLCA [52]. According to recent studies, Keratin 17 (K17) targets tumour suppressors and mediates signals to promote tumour growth [53]. K17 is not only a very promising therapeutic target but is also used to improve the diagnostic accuracy of UC and detect recurrence [53,54]. Our study suggests that FBN2 may affect cornification, but the specific regulatory mechanism still needs further study.
Natural killer (NK) cells are important innate immune cells to kill cancer cells without prior sensitisation [55,56]. The results of our Immune cell infiltration analysis suggested that high FBN2 may lead to NK cells resting and the reason may be related to immunosuppression of the TME (55). According to the study, NK cells at rest are not hazardous to tumour cells, however NK cells that have been activated by IL-2 or IL-15 have the ability to prevent tumour growth [57]. Mast cell (MC) has interactions with immunosuppressive cells and is associated with inflammatory [58,59]. MC plays complex and dynamic functions in different cancers, which may promote or inhibit cancer [59]. The macrophages in TME are called the TAMs which are related to poor prognosis [60]. The high FBN2 expression group had significantly higher expression of immunosuppressive molecules (e.g. CTLA-4) and increased levels of various immunosuppressive cells (e.g. macrophages) than the low FBN2 expression group, suggesting that the weakened immune phenotype in the high FBN2 expression group may be due to its stronger immunosuppressive environment and immune checkpoint expression than the low FBN2 expression group. At present, the understanding of the role of TAM in TME in BLCA is limited and it is still an is a research hotspot due to its important role in the progression and metastasis of BLCA [61]. This study suggests that FBN2 may regulate the immune microenvironment through the abovementioned immune cells, thereby affecting the progression of bladder cancer.
In our study, patients with high levels of FBN2 expression also had greater amounts of TMB, and these two factors worked together to predict how patients with BLCA would do. A potential predictive biomarker of clinical response to immune checkpoint inhibitor treatment is TMB, or the amount of somatic missense mutations per million bases (MB) in a tumour's genes [62]. Neoantigens are produced as a result of the slow accumulation of somatic mutations, which activate T cell immunogenicity and suppress tumour cells [63]. High TMB tumours produce many antigens, making them immunogenic and triggering an anticancer response [64]. Our findings demonstrated that increased TMB tended to predict poorer survival, which was consistent with the earlier investigations.
In conclusion, we analysed the relationship between FBN2 expression and BLCA and demonstrated that FBN2 can serve as an independent prognostic factor for BLCA. We also provide F I G U R E 6 Box plots of the relationship between FBN2 and drug susceptibility of Cisplatin(A, p = 0.0004), Docetaxel(B, p = 0.00028), Doxorubicin(C, p = 0.00015), Gemcitabine(D, p = 0.0005), Methotrexate(E, p = 0.24), and Vinblastine(F, p = 0.023). The lower the IC50 is, the higher the drug sensitivity. The number between the two boxes is the p value.

170
new insights into the relationship between the TME and FBN2 in BLCA. Therapies targeting FBN2 may help mitigate immune escape and enhance the effectiveness of immunotherapy.