The RNA sequencing data (FPKM value), somatic mutation data, and clinicopathological features for various cancer types of The Cancer Genome Atlas (TCGA) were obtained from the UCSC Xena (https://xenabrowser.net/datapages/) data portal (Goldman et al., 2020). FPKM values were converted into TPM values using the following formula: T P M i = F P K M i Σ j F P K M j × 10 6

Following that, we applied a log2^(TPM+1) transformation to the TPM value for further investigation. The somatic mutation data were processes via Varscan (https://varscan.sourceforge.net/) (Koboldt et al., 2013). The data on copy number variation (CNV) of the genes that indicate hyperprogression were obtained from a previous research (Hu et al., 2021). Additionally, for external validation purposes, we downloaded three independent cohorts namely GSE48075,E-MTAB-4321 and IMvigor210 (http://research-pub.gene.com/IMvigor210CoreBiologies/) (Choi et al., 2014; Hedegaard et al., 2016; Mariathasan et al., 2018). And we collected six immunotherapy cohorts in melanoma. These datasets were also converted from FPKM to TPM values for subsequent analysis.

The immunological features of TME in BLCA comprise the presence of immunomodulators, the functioning of the cancer immunity cycle, the degree of infiltration by tumor infiltrating immune cells (TIICs), and the presence of inhibitory immune checkpoints. Initially, data on 121 immunomodulators, encompassing major histocompatibility complex (MHC), receptors, chemokines, and immunostimulators, were gathered from Charoentong et al.'s research (Charoentong et al., 2017). The cancer immunity cycle represents the immune response against cancer and comprises of seven stages: release of cancer cell antigens (step 1), cancer antigen presentation (step 2), priming and activation (step 3), trafficking of T cells to tumors (step 4), infiltration of T cells into tumors (step 5), recognition of cancer cells by T cells (step 6), and killing of cancer cells (step 7) (Chen and Mellman, 2013). Investigation of these steps' activity was conducted through the utilization of a single sample gene set enrichment analysis (ssGSEA).

Using the R package “GSVA,” we performed ssGSEA to assess the extent of TIIC infiltration in the BLCA, utilizing signatures from the TISIDB database (Ru et al., 2019). Then, to avoid bias, six independent algorithms were used to determine the level of TIIC infiltration in BLCA, namely, TIMER (Li et al., 2017), MCP-counter (Becht et al., 2016), CIBERSORT (Newman et al., 2015), quanTIseq (Finotello et al., 2019), xCell (Aran et al., 2017), and EPIC (Racle et al., 2017). From previous studies, we adopted effector genes of TIICs as well (Hu et al., 2021). Next, we gathered several inhibitory immune checkpoints with promising therapeutic potential as reported in Auslander’s research (Auslander et al., 2018).

T cell inflammation score (TIS), was employed for assessing the inflammation level within the TME in BLCA. TIS serves as a previously developed predictive factor for assessing the effectiveness of cancer immunotherapy and anti-PD-1 therapy (Ayers et al., 2017). Calculating the TIS, which includes 18 IFN-γ-responsive genes, helps to reflect pre-existing anti-cancer immunity and predict the clinical response to ICB. Genes collected in this study were identified by Ayers et al. (Ayers et al., 2017). T I S = Σ γ = 1 18 β γ X γ where βγ represents the coefficient of a gene predefined in a previous study, and Xγ is the expression level of this gene.

Hyperprogression is considered to be an adverse event of abnormally accelerated tumor growth when ICB is performed. We identified several predictors used to predict hyperprogression from previous studies (Kato et al., 2017; Singavi et al., 2017; Giusti et al., 2019). Kato et al. reported in Clinical Cancer Research on a study investigating the association between genomic variants and hyperprogression. Among 155 patients with tumors treated with anti-PD-1/PD-L1 monotherapy, the study revealed that six patients with MDM2/MDM4 and DNMT3A amplification experienced TTF of less than 2 months. Singavi found that patients with chromosome 11 region 13 amplification variants (CCND1, FGF3, FGF4, and FGF19 amplification) were prone to hyperprogression on immunotherapy by examining the occurrence of somatic mutations in 696 patients with solid tumors. Giusti et al. collected clinical data on 20 patients with advanced NSCLC treated with pembrolizumab immunotherapy, in which five hyperprogressing patients were identified, and NGS revealed CDKN2A/B deletion in 4/5 hyperprogressing patients.

CD8 and PD-L1 staining was conducted using the BLCA tumor tissue microarray (HBlaU050CS01), which included 40 bladder cancer tissue samples and 10 adjacent peritumoral tissues, offered by Shanghai Outdo Biotech Company located in Shanghai, China. The positive ratio of CD8⁺ T cells were defined based on a comparison of the infiltration ratio of CD8⁺ cells within each nest to the number of total cells within each nest. Only the proportion of strongly positive cells were recorded, while the proportion of weakly positive cells was disregarded. The ethics committee of Shanghai Outdo Biotech Company granted approval for this study. Immunohistochemistry of the tumor tissue microarray was performed by Biossci Company in Hubei, China. An antibody against IL4I1 (ab222102) was purchased from Abcam company. For IL4I1 and PD-L1, we performed a semi-quantitative evaluation for staining intensity, categorizing it as negative (0), weakly positive (1⁺), moderately positive (2⁺), or strongly positive (3⁺), and determining the percentage of positive cells present. To determine the histochemistry score (H-score) for each observed tissue component (cytoplasm and nucleus), we multiplied the intensity score (ranging from 0 to 3) and the percentage of positive cells (ranging from 0 to 100).

To analyze the molecular subtypes in the samples and establish the correlation between IL4I1 expression and molecular subtypes, we utilized a range of subtype systems such as CIT, Lund, MDA, TCGA, Baylor, UNC, and Consensus subtypes and determined each sample’s subtype using consensusMIBC and BLCAsubtyping R packages (Damrauer et al., 2014; Rebouissou et al., 2014; Robertson et al., 2017; Marzouka et al., 2018; Mo et al., 2018; Kamoun et al., 2020). The molecular subtype of BLCA can be classified into binary subtype, including basal subtype and luminal subtype (Kamoun et al., 2020). The Basal subtype, which is considered more aggressive, also shows an increased response to certain treatments such as immunotherapy and anti-EGFR therapy. We also have gathered a total of 12 specific bladder cancer gene signatures (Kamoun et al., 2020). To assess the predictive precision of IL4I1 in identifying BLCA molecular subtypes, we created receiver operating characteristic (ROC) curves.

A set of gene signatures positively correlated with clinical response to anti-PD-L1 (atezolizumab) in BLCA from Mariathasan’s study (Mariathasan et al., 2018). Significantly, the determining factors for neoadjuvant chemotherapy in BLCA encompass the genetic mutation status of various pivotal genes like RB1, ATM, ERBB2, ERCC2, and FANCC (van Allen et al., 2014; Groenendijk et al., 2016; Pietzak et al., 2019; Singla et al., 2019). Additionally, we gathered additional therapeutic signatures, such as oncogenic pathways that could contribute to a non-inflamed TME, the EGFR ligands and radiotherapy predicted pathways. We used the ‘GSVA’ R package to calculate the enrichment scores for these signatures (Hänzelmann et al., 2013). Lastly, we acquired target genes of drugs employed in various therapeutic approaches by Drugbank (Wishart et al., 2018) database (https://www.drugbank.com/) and examined their association with IL4I1.

To evaluate differences among continuous variables, either the Wilcoxon ranking test or Kruskal Wallis test was utilized, depending on the quantity of groups, whereas the chi-square test was employed to examine distinctions among categorical variables. The Pearson correlation analysis was utilized to quantify correlation. The survival cruve were generated using the Kaplan-Meier method and the prognostic value were obtained via univariate Cox regression. Statistical analyses were all conducted using R project (version 4.2.3). A significance level of 0.05 was used, and p-values less than this threshold were considered statistically significant.

Firstly, 1863 differential expressed genes between normal and tumor samples in the TCGA cohort were identified, followed by 92 DEGs between inflamed tumor samples and non-inflamed tumor samples in the IMvigor210 cohort. After crossing the DEGs from the two cohorts, we finally identified 18 common DEGs (Figure 2A). Figure 2B demonstrated the relative expression of these 18 DEGs. For these genes, IL4I1, CXCL9, CXCL10, CXCL11, KLHDC7B, and TNFRSF18 had a relative high expression in tumor samples. Here, we focused on the role of IL4I1 expression in bladder cancer. Then, we explored the pan-cancer expression pattern of IL4I1 by analyzing the different expression between normal and tumor tissues in 23 tumor types. It was found that the expression of IL4I1 was significantly higher in tumor tissues than normal tissues in most tumor types such as BLCA, HNSC, PRAD, KIRP, and so on (Figure 2C). The paired analysis showed the same trend (Figure 2D). Subsequently, we performed GSEA for IL4I1 in terms of Hallmark gene sets or KEGG gene sets based on the TCGA-BLCA cohort. As demonstrated, IL4I1 was significantly positively correlated with immune-related pathways, including IL-2-STAT5 signaling, IL4-JAK-STAT3 signaling and antigen processing and presentation (Figures 2E, F).

We have demonstrated that IL4I1 is highly expressed in a variety of tumor tissues, prompting us to explore the prognostic value of IL4I1. We assessed the prognostic value of IL4I1 through univariate cox analysis and survival curves. Although IL4I1 has no prognostic value in bladder cancer, its overexpression was observed as a significant risk factor for ACC, GBM, KIRC, KIRP, LAML, LGG, LIHC, THYM, and UVM in terms of overall survival (OS), while a protective effect was observed only in SKCM (Supplementary Figure S1). The analysis of progression-free survival (PFS) revealed that high expression of IL4I1 was also found to be a contributing factor for GBM, KIRCK, IRPK, and LGG. In addition to SKCM, IL4I1 overexpression also had a significant protective effect on CESC, CHOL, and HNSC (Supplementary Figure S2). However, in the disease-free survival (DFS) analysis, only KIRP and THCA had a significant risk effect and were protective factors for OV (Supplementary Figure S3). Considering the results of GSEA suggests that IL4I1 is closely related to immune-related pathways, a comprehensive analysis was performed to identify the correlation between IL4I1 and immunological characteristics across various types of tumors. It was found that the expression of IL4I1 showed positive correlation with the vast majority of immunomodulators in almost all tumors, especially BLCA, HNSC, and KICH (Figure 3A). However, in certain tumors, such as THCT and UCEC, it is negatively correlated with some immunomodulators. In multiple types of cancers, we discovered a positive correlation between IL4I1 expression and various immune checkpoints, including PD-1, CTLA-4, PD-L1, and Tim-3 (Figures 3B–E). Next, the ssGSEA algorithm was utilized to analyze the infiltration level of TIICs. In contrast, except for DLBC and CHOL, IL4I1 showed a positive association with various types of TIICs in different tumors (Figure 3F). These findings suggested that the expression of IL4I1 holds immunological relevance and prognostic value across various types of tumors, suggesting that IL4I1 can serve as a target for immunotherapy.

Subsequently, we proceeded to analyze the immunological properties of IL4I1 in BLCA. Given that the immune system is a complex and interconnected network involving numerous molecules and pathways. To further evaluate the immunological functions of IL4I1, we assessed its impact on the cancer immune cycle. Our findings indicated that the enrichment score of most steps in the cancer immune cycle is upregulated in the high IL4I1 group (Figure 4A). The process involved release of cancer cell antigens (step 1), priming and activation (step 3), as well as trafficking of T cells to tumors (step 4) (except for B cell recruitment, all other steps including CD8⁺ T cell, DC, macrophage, Th1 cell, NK cell, MDSC, and Th17 cell recruitment showed increased activity). The connection between these processes and the infiltration of immune cells in the TME is evident through the heightened T cell infiltration into tumors (step 5) observed in the high-IL4I1 group. Furthermore, there was an upregulation in the activity of eliminating cancer cells (step 7) within the high-IL4I1 group. Significantly, the T cells' ability to identify cancer cells (step 6) was observed to be reduced in the high IL4I1 group, possibly because of the increased inhibitory immune checkpoints expression in this specific group.

Furthermore, we employed the TIMER and MCP-counter algorithms to quantify the infiltration levels of immune cells and analyzed their relationship with IL4I1. The results displayed a favorable association between IL4I1 manifestation and the levels of infiltration by CD8⁺T cells, macrophages, and dendritic cells (Figures 4B, C). Additionally, the expression of IL4I1 exhibited a positive relationship with the effector genes of TIICs (Figure 4D). Furthermore, we utilized six different algorithms to further investigate the relationship between IL4I1 and TIICs (including CD8⁺ T cells, NK cells, Th1 cells, dendritic cells, and macrophages) (Figure 4E).

To summarize, these findings indicated a notable correlation between the excessive expression of IL4I1 and the inflamed TME within BLCA.

Based on the previous description, we proposed that IL4I1 suggested an inflamed TME, thereby suggesting that high expression of IL4I1 may confer increased sensitivity to ICB therapy. Consequently, we embarked on an analysis to explore the association between IL4I1 and 20 inhibitory immune checkpoints. With the exception of LGALS3 and VTCN1, the expression levels of immune checkpoints exhibited significant upregulation. Correlation analysis further indicated a positive association between IL4I1 expression and these immune checkpoints (Figures 5A, D). Furthermore, IL4I1 exhibited positive correlations with certain pathway enrichment scores that are known to be favorable for immunotherapy (Figure 5B). Furthermore, there is a significant and positive association between IL4I1 and TIS (Figure 5C). The above results suggested that IL4I1 has the capacity to serve as a promising indicator for the effectiveness of ICB. Conversely, the high-IL4I1 group exhibited a lower incidence of hyperprogression. The correlation between IL4I1 and hyperprogression related genes was examined, and our results showed a notable decrease in the expression of genes that exhibited a positive correlation with the occurrence of hyperprogression in the high IL4I1 group. This reduction was observed in genes such as MDM2, MDM4, as well as a significant decrease in the rate of copy number amplification, including CCND1, FGF19, FGF3, and FGF4 (Figures 5E, F). Conversely, two predictors that exhibited a negative correlation with hyperprogression, namely CDKN2A and CDKN2B, exhibited significantly higher expression levels in the high-IL4I1 group compared to the low-IL4I1 group, along with a higher copy number amplification rate. These findings suggested that patients exhibiting high IL4I1 expression are less prone to experiencing hyperprogression associated with ICB treatment, thus making them more appropriate candidates for ICB therapy.

The different molecular subtypes of BLCA affect the prognosis and treatment effect of patients. Hence, an analysis was conducted to examine the correlation between the IL4I1 expression and molecular subtypes across various classification systems. Figure 6A demonstrated that the high-IL4I1 group displayed a stronger preference for the basal subtype of BLCA in the findings. Furthermore, the low-IL4I1 group exhibited elevated enrichment scores for urothelial differentiation, the Ta pathway, and luminal differentiation (Figure 6A). The high-IL4I1 group exhibited elevated enrichment scores in basal differentiation, EMT differentiation, immune differentiation, smooth muscle, myofibroblast, interferon response, and keratinization. Moreover, apart from the Baylor system, all molecular subtypes exhibited an AUC value exceeding 0.7 (Figure 6B), signifying that IL4I1 has the ability to accurately anticipate the molecular subtype of BLCA, thus providing precise treatment guidance.

In addition to immunotherapy, IL4I1 can predict the efficacy of other therapeutic strategies, like neoadjuvant chemotherapy, radiotherapy, and various targeted therapies. We collected mutation data for molecules associated with neoadjuvant chemotherapy, various therapeutic signatures, and gene expression related to targeted drugs. In the high-IL4I1 group, the mutation frequency of specific genes (RB1, ERBB2, ERCC2) linked to neoadjuvant chemotherapy exhibited a higher rate (Figures 6C, D). Furthermore, the group with high IL4I1 levels demonstrated increased enrichment scores for pathways linked to EGFR ligands and radiotherapy. On the contrary, the group exhibiting low IL4I1 levels demonstrated an enrichment of oncogenic pathways, as depicted in Figure 6E. The activation of these pathways was found to be connected with a non-inflamed TME and resistance to immunotherapy (Peng et al., 2015; Sweis et al., 2016; Spranger and Gajewski, 2018). By utilizing the Drugbank database, we discovered an observation indicating significantly higher expression levels of numerous drug target genes in the high-IL4I1 group (Figure 6F). In particular, the high IL4I1 group exhibited a significant increase in nearly all aspects related to chemotherapy, immunotherapy, and ERBB target gene expression. And some target genes of anti-angiogenesis (including SH2B3, FTL3, FTL4, and CDF1R) were also significantly increased, while the expression levels of BRAF and RAF1 were low.

In general, these results suggested that BLCA patients with high IL4I1 expression may benefit more from neoadjuvant and adjuvant chemotherapy, radiotherapy, immunotherapy, and ERBB therapy.

To provide more robust results, we performed a validation analysis in the GSE48075 cohort and E-MTAB-4321 cohort. There was a significant increase in the expression of the majority of immune checkpoints in patients with high IL4I1 expression levels (Figure 7A; Supplementary Figure S4A), and correlation analysis also showed a positive correlation (Figure 7B). Similarly, the expression of IL4I1 was positively correlated with TIS (Figure 7C). Several immunotherapy positive predictable pathways also significantly enriched in high-IL4I1 group (Figure 7D, Supplementary Figure S4B) and are positively associated with IL4I1 (Figure 7E). Furthermore, the majority of effector genes in TIICs exhibited a positive correlation with IL4I1, as depicted in Figures 7F, G; Supplementary Figure S4C. Two external cohort revealed a significant correlation between elevated IL4I1 expression and the basal molecular subtypes of BLCA, thereby validating IL4I1’s potential as a predictive indicator for BLCA molecular subtypes (Figures 8A, B; Supplementary Figures S4D, E). For several therapeutic strategies, IL4I1 also showed its predictive value. The high-IL4I1 group showed significant enrichment of EGFR and radiotherapy predictive pathways (Figure 8C; Supplementary Figure S4F), and IL4I1 was positively correlated with the expression of drug target for chemotherapy, immunotherapy, ERBB and anti-angiogenic therapy (Figure 8D). The validation of the GSE48075 and E-MTAB-4321 cohort further confirmed IL4I1 suggested an inflamed TME and the potential of IL4I1 to predict molecular subtypes and several therapeutic strategies for BLCA.

Next, we performed immunohistochemical staining of BLCA and classified the tumor tissue microarray cohort into inflamed, excluded, and desert subtypes (Ott et al., 2019) derived from the spatial distribution pattern of CD8⁺ T cells (Figure 9A). In the inflamed subtype, the H-score of IL4I1 and PD-L1 were found to be higher comparison to the other two subtypes (Figures 9B, C). Figures 9D–F showed that there was a positive association between the H-score of PD-L1 and the rate of CD8 positivity, while the H-score of IL4I1 was positively linked to the H-score of PD-L1 and the positive rate of CD8. In the IMvigor210 cohort (a BLCA immunotherapy cohort), according to the clinical efficacy of the patients, the cohort was divided into PR/CR group and PD/SD group, and the correlation of IL4I1 with immunological characteristics and its predictive value for molecular subtypes and therapy were explored respectively. And we also arrived a similar conclusion. (Supplementary Figures S6, S7). Additionally, we conducted investigations into the involvement of IL4I1 in immunotherapy. As anticipated, the high-IL4I1 group displayed notably elevated percentages of IC2 (immune cells with the utmost PD-L1 level) and TC2 (tumor cells with the utmost PD-L1 level) (Figure 9G), aligning with our expectations for patients who received anti-PD-1 therapy. In addition, the high-IL4I1 group had a higher survival rate and response rate to anti-PD-L1 therapy, and IL4I1 expression was higher in the complete response group (Figures 9H–J). These findings collectively suggested that IL4I1 was closely related to the formation of inflamed TME and has a certain value in predicting the response to immunotherapy in bladder cancer. Additionally, we collected six immunotherapy cohorts, including five melanoma ICB treatment cohorts and one adoptive T cell therapy cohort to investigate the predictive role of IL4I1 in immunotherapy response. Although not reaching statistical significance, it is noteworthy that in cohorts such as Gide 2019, Nathanson2017 post, and GSE78220, the high IL4I1 group exhibited higher response rates and thus a more favorable prognosis, consistent with our findings in bladder cancer. In contrast, in cohort Nathanson2017 pre, the high IL4I1 group displayed a lower response rate compared to the low IL4I1 group, resulting in a poorer prognosis (Supplementary Figures S8A–F).