2.1 Ethics Statement

This clinical research scheme was authorized by the Meizhou People's Hospital Clinical Scientific Research and New Technology Ethics Committee (Huangtang Hospital, Meizhoulunshen 2021-C-29, July 22, 2021). It was performed in accordance with the Declaration of Helsinki (1975) and the International Ethical Guidelines for Biomedical Research Involving Human Subjects (2002). Before specimen collection, written informed consent was obtained from all participants.

2.2 Patients, Clinical Samples, and Data

Overall, 20 patients histopathologically diagnosed with NMIBC at the Meizhou Peoplés Hospital (Huangtang Hospital) from March 2017 to May 2021 were included in this study for RNA-sequencing. Additionally, 43 patients from June 2021 to February 2024 were included for the external validation cohort. Patients (1) who were diagnosed with NMIBC by histopathological examination; (2) who were initially diagnosed with and treated for NMIBC, but with no history of radiotherapy and chemotherapy; (3) with complete clinicopathological and follow-up data; and (4) who signed an informed consent form were included in the study. In contrast, patients (1) who developed complications with other malignant tumors; (2) whose NMIBC was complicated with acute cystitis, urethral stricture, and abnormal coagulation function; and (3) developed severe liver and kidney dysfunction.

Data on age, gender, tumor node metastasis (TNM) stage, histologic grade, tumor number, tumor size, and degree of differentiation of all included patients were recorded. A total of 20 NMIBC patients were analyzed; of them, seven patients experienced relapse, while 13 patients did not experience relapse after HIPEC treatment. Table 1 presents the baseline clinical features of the relapsed group and non-relapsed group. After TURBt treatment, tumor tissues were collected from the patients with NMIBC, respectively, during surgery and immediately placed in liquid nitrogen. The clinical specimens were then stored in a −80 °C refrigerator for subsequent RNA-Seq. One sample will be divided into two parts: one for RNA-Seq and the other for rt-qPCR. The specific processing flow is shown in Figure 1.

2.3 TURBt and HIPEC Treatment

After the initial diagnosis, all patients underwent TURBt following the standard protocols. Within 1 week after the surgical procedure, appropriate chemotherapeutic agents were selected for intravesical hyperthermic chemotherapy. High-risk NMIBC patients who have undergone TURBt received hyperthermic intracavitary perfusion treatment using the BR-TRG-II system. The perfusion system is configured to operate in a circulation mode, maintaining the inflow temperature at 45.0°C ± 0.3°C and the outflow temperature at 43.0°C ± 0.5°C, while regulating the perfusion rate between 100 and 200 mL/min. The perfusion rate may be adjusted based on individual patient tolerance, maintaining an appropriate perfusion pressure to ensure optimal bladder filling. The duration of hyperthermic perfusion treatment approximates 45 min. Intravesical hyperthermic chemotherapy with PiRubicin was administered at a dose of 15–30 mg/m², adjusted according to body surface area, and diluted to a concentration of 500 μg/mL. Treatment was continuously administered for 12 months, once a week during the initial 2 months and once a month for another 10 months. In patients with tumor recurrence, a repeat TURBt was performed, and the excised tissue was utilized for RNA-seq of the recurrent NMIBC tissue.

2.4 Total RNA Isolation, Library Preparation, and Sequencing

Fourteen samples from the relapsed group (seven for pre-HIPEC and seven for post-HIPEC) and 13 samples from the non-relapsed group were submitted for RNA-Seq. The total RNA was extracted using the Qiagen RNeasy formalin-fixed paraffin-embedded kit (73,504, Qiagen, Hilden, Germany) in accordance with the manufacturer's protocol. The purity and quantity of the obtained total RNA were determined using the NanoDrop spectrophotometer. The integrity of the RNA was analyzed using the RNA Nano6000 Assay Kit on the Bioanalyzer 2100 system (Agilent Technologies, CA, USA). Approximately 1 μg of RNA per sample was utilized as input for the subsequent RNA sample preparations. Strand-specific RNA-Seq libraries were then generated utilizing the Whole RNA-seq Lib Prep kit for Illumina (RK20303, ABclonal, Shanghai, China). The quality of the resulting libraries was evaluated using the Agilent Bioanalyzer 2100 system (Agilent, USA). Ultimately, the prepared libraries were subjected to sequencing at the Novogene Bioinformatics Institute (Beijing, China) on an Illumina HiseqX10 platform using a 150-bp paired-end read protocol.

2.5 Gene Expression Quantification

The initial RNA-Seq reads underwent filtering using FastQC. Subsequently, the reads were aligned to the Ensembl human genome assembly using STAR (v2.7.0 f) with default parameters. The gene expression levels were measured using both raw count and transcripts per kilobase million. The annotation information of the mRNA within the human genome was retrieved from the GENCODE (v19) database.

2.6 Analysis of Differentially Expressed Genes (DEGs) and Enrichment of Signaling Pathways

Three groups of patients underwent differential expression analysis: (1) seven with relapse NMIBC (pre-HIPEC) and 13 with non-relapse NMIBC and (2) seven with pre-HIPEC and seven post-HIPEC. The R package “DESeq2” was used for determining the expression status of DEGs based on a cutoff p ≤ 0.05 and |a log2FoldChange| of ≥ 1. The upregulated genes and downregulated genes were imported into the “clusterProfiler” software for ontology and pathway enrichment analysis utilizing the Gene Ontology and KEGG databases.

2.7 Calculation of the Immune and Stromal Scores and Estimation

The Estimation of STromal and Immune cells in MAlignant Tumor tissues using Expression (ESTIMATE) algorithm is a computational approach used to estimate the fraction of tumor purity in malignant tissues by leveraging the expression data of immune cells and stromal cells. In this study, we employed the ESTIMATE algorithm to assess the immune grade, stromal grade, and tumor purity of patients with NMIBC, considering the Immunity_H, Immunity_M, and Immunity_L groups as reference points. To further investigate the composition of immune cell types within each subset, we utilized the CIBERSORT (available at https://cibersortx.stanford.edu/), EPIC, and quanTIseq method, a biological approach for cell-type identification by estimating relative subsets of RNA transcripts. The CIBERSORT package was employed to calculate the distribution of immune cell types within each subset. A comparison of immune cell proportions among NMIBC subtypes was performed using the Kruskal–Wallis test. The ESTIMATE and CIBERSORT packages in R (version 3.6.2, available at https://www.R-project.org/) were utilized for the analyses.

2.8 qRT-PCR

The total RNA was extracted from cells with TRIzol reagent (Invitrogen, Carlsbad, CA, USA). cDNA was generated according to the manufacturer's protocol (Takara). Quantitative reverse transcription polymerase chain reaction (qRT-PCR) was performed using an Applied Biosystems instrument at a reaction volume of 20 μL (SYBR Green Real-Time PCR Master Mix; Tsingke, Beijing, China). The gene GAPDH was used as an internal parameter [12]. All samples were tested in triplicate. qRT-PCR primers for this study are listed in Table S1.

2.9 Statistics

All analyses were conducted using the R environment (v3.6.0). All comparisons for continuous variables were performed using the two-sided Mann–Whitney U test for two groups, and the differential expression gene (DEG) comparisons between pre- and post-HIPEC treatment groups used a paired test design. TMEFF2, KRT222, and GTSF1 were evaluated for predicting relapse in NMIBC patients after HIPEC treatment, while ZMAP4, UPP2, and GALR1 were assessed for predicting relapse in NMIBC patients before HIPEC treatment through ROC curve analysis. A p < 0.05 was considered significant. The Benjamini–Hochberg method was used for multiple testing correction.

3.1 RNA-Sequencing Reveals DEGs of in Relapsed NMIBC Patients After HIPEC Treatment

There is no significant difference in clinical baseline characteristics between the relapsed and non-relapsed groups (Table 1). To determine the underlying mechanism of NMIBC relapse in patients who underwent HIPEC treatment, we compared the expression status of DEGs in the high-risk NMIBC pre- and post-HIPEC treatment groups. A total of 1277 DEGs (p < 0.05) were identified in NMIBC patients who experienced relapse before and after HIPEC treatment. The top 20 differentially expressed markers are shown in Table 2. Furthermore, volcano plots and heatmaps were generated to visually represent the DEGs detected in the pre-HIPEC and post-HIPEC groups (Figure 2A,C). After HIPEC treatment, the NMIBC tissue had higher mRNA expression levels of TMEFF2, KRT222, IGKV6D−21, GTSF1, IGHGP, AL592429.2, HOXC10, ZIC1, GABRA2, and HMGB1P35.

Moreover, to identify the genetic characteristics in relapsed NMIBC patients after HIPEC treatment, a comparison of the DEGs was conducted between the pre-HIPEC and non-relapsed NMIBC groups. A total of 1158 DEGs (p < 0.05) were identified in the relapsed and non-relapsed NMIBC patients before HIPEC treatment. The top 20 differentially expressed markers are shown in Table 3. Relapsed NMIBC patients (pre-HIPEC) exhibited higher expression levels of IGHV1OR15–1, AC135068.8, ZMAT4, AC110792.2, UPP2, AL844908.1, ENSG00000179449.2, GALR1, REXO1L12P, and FAM71C compared with non-relapsed NMIBC patients (Figure 2B,D). These findings depict the transcriptomic profiles of relapsed NMIBC patients before and after HIPEC treatment.

Finally, we employed qRT-PCR to validate the expression of selected markers. We observed a significant increase in the expression levels of TMEFF2, KRT222, and GTSF1 after HIPEC treatment in the relapse cohort, suggesting that HIPEC treatment mediates the upregulation of these genes (Figure 2E). Additionally, ZMAP4, UPP2, and GALR1 showed significantly higher expression in the pre-HIPEC patients in the relapse cohort compared with the non-relapsed patients (Figure 2F). These findings indicate potential markers for predicting the response of HIPEC treatment in NMIBC patients. Moreover, these results affirm the consistency of the RNA-seq and qRT-PCR findings.

3.2 GO and KEGG Analysis of DEGs in Relapsed NMIBC Patients

To elucidate the functional implications of differentially expressed genes (DEGs) in relapsed NMIBC samples, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses were conducted on the identified DEGs. In the context of relapsed NMIBC samples, the DEGs identified before and after HIPEC treatment were predominantly enriched in various biological processes, including cell recognition, immunoglobulin-mediated immune response, complement activation, phagocytosis, and others. The enrichment analysis of cellular components and molecular functions is depicted in Figure 3A. In relapsed and non-relapsed NMIBC samples, the DEGs were primarily enriched in developmental maturation, the activation of the adenylate cyclase-modulating G protein-coupled receptor signaling pathway, and the regulation of signaling receptor activity. The cellular component and molecular function enrichments are illustrated in Figure 3B.

In the KEGG enrichment analysis, the DEGs identified in relapsed NMIBC samples before and after HIPEC treatment were associated with various pathways, including the Wnt signaling pathway, cell adhesion molecules, type I diabetes mellitus, and the intestinal immune network for IgA production (Figure 3C). Moreover, the DEGs distinguishing relapsed from non-relapsed NMIBC samples were implicated in cytokine–cytokine receptor interaction, neuroactive ligand–receptor interaction, and the Wnt signaling pathway (Figure 3D). Collectively, these findings provide robust evidence supporting the pivotal role of the Wnt signaling pathway as a crucial molecular characteristic in relapsed NMIBC samples.

3.3 Immunoscore Analysis in Relapsed and Non-Relapsed NMIBC Samples

To further examine the changes in the tumor microenvironment (TME) within the samples of relapsed and non-relapsed NMIBC patients, we conducted an ESTIMATE analysis of different samples. The NMIBC samples after HIPEC treatment showed a higher Stromal Score compared with the non-relapsed NMIBC samples and the relapsed NMIBC samples before HIPEC treatment. However, no significant differences were observed in the total immune scores and ESTIMATE score between non-relapsed, pre-HIPEC, and post-HIPEC relapsed NMIBC samples. Results of the CIBERSORT analysis of immune cell scores in different groups are shown in Figure 4D.

3.4 Comparison of Immune Infiltration Levels in Relapsed and Non-Relapsed NMIBC Samples

To elucidate the alterations in the tumor microenvironment (TME) within relapsed NMIBC samples, we investigated immune infiltration levels in both relapsed and non-relapsed NMIBC samples using CIBERSORT analysis. No significant differences were observed in the infiltration levels of activated memory CD4 T cells; resting memory CD4 T cells; native CD4 T cells; CD8 T cells; memory B cells; resting dendritic cells; M0 macrophages; M1 macrophages; M2 macrophages; activated mast cells; resting mast cells; eosinophils; neutrophils; activated natural killer (NK) cells; resting NK cells; plasma cells; and regulatory T cells (Tregs) (Figure 5A–E,H,J–T).

However, notable differences were observed in immune cell infiltration between relapsed NMIBC samples after HIPEC treatment. Specifically, these samples exhibited higher levels of native B-cell infiltration (Figure 5F), along with lower levels of infiltrated activated dendritic cells (Figure 5G) and monocytes (Figure 5I). Moreover, we also perform the immune infiltration between the three groups in Figure S2 via EPIC and Figure S3 via quanTIseq method. We found the post-HIPEC group has higher endothelial (Figure S2) and M2 macrophages (Figure S3). These findings provide further evidence of the dynamic changes occurring within the TME of relapsed NMIBC samples before and after HIPEC treatment.

3.5 Validation in the Validation Cohorts

Finally, we further collected a validation cohort that included 43 NMIBC patients. Of these, 16 patients experienced relapse, while 27 patients did not experience relapse after HIPEC treatment. Table 4 presents the baseline clinical features of the relapsed and non-relapsed groups. We found a significant increase in the expression levels of TMEFF2, KRT222, and GTSF1 after HIPEC treatment in the relapse cohort (Figure 6A). The area under the curve (AUC) for predicting relapse in NMIBC patients after HIPEC treatment was as follows: TMEFF2: AUC = 0.956 (sensitivity 96.3%, specificity 77.78%, and cutoff point 1.875); KRT222: AUC = 0.842 (sensitivity 93.2%, specificity 48.15%, and cutoff point 1.245); GTSF1: AUC = 0.922 (sensitivity 94.7%, specificity 48.15%, and cutoff point 1.555) (Figure 6B).

Additionally, ZMAP4, UPP2, and GALR1 showed significantly higher expression in pre-HIPEC patients in the relapse cohort compared with the non-relapsed patients (Figure 6C). The AUC for predicting relapse in NMIBC patients before HIPEC treatment was as follows: ZMAP4: AUC = 0.944 (sensitivity 95.8%, specificity 55.56%, and cutoff point 1.443); UPP2: AUC = 0.798 (sensitivity 92.6%, specificity 51.85%, and cutoff point 1.345); GALR1: AUC = 0.857 (sensitivity 92.3%, specificity 59.26%, and cutoff point 1.706) (Figure 6B).