Herein, clinical data and expression profiles of 33 forms of cancers were obtained from The Cancer Genome Atlas (TCGA) (https://portal.gdc.cancer.gov/). Additionally, we collected the TCGA and healthy sample data from the TCGA_ Genotype-Tissue Expression (GTEx) dataset, available at the University of California, Santa Cruz (UCSC) Xena (https://xenabrowser.net/datapages/). Furthermore, the healthy and malignant tissue samples were retrieved based on the immunohistochemical image data derived from Human Protein Atlas (HPA) (https://www.proteinatlas.org/). Furthermore, RBM4 and Gene Expression Profiling Interactive Analysis Database (GEPIA2) (http://gepia2.cancer-pku.cn/#index) were searched for extracting the Top 100 relevant genes from the TCGA dataset. All experiments in the study were based on the standard guidelines issued by TCGA and UCSC. Therefore, we did not acquire ethical approval or informed consent from the patients.

The RBM4 mRNA expression levels in the normal and malignant tissues acquired from the TCGA, TCGA_GTEx, and TCGA paired samples were compared in this study. Furthermore, the RBM4 protein expression levels measured in the data acquired from the healthy and cancerous tissues in the HPA database were also investigated.

Kaplan-Meier (KM) analysis and the log-rank test were conducted to determine the relationship between RBM4 expression levels and different clinical outcomes, such as progression-free interval (PFI), overall survival (OS), and disease-specific survival (DSS) in pan-cancer using TCGA. We displayed survival curves with a significant p-value 0.05. Furthermore, receiver operating characteristic (ROC) curves were constructed for the tumors where RBM4 expression could potentially affect prognosis.

This study has examined the correlation between RBM4 expression levels and other clinical factors, like T stage, gender, N stage, and pathologic stage, in the case of cancers that are affected by RBM4 and can influence prognosis.

Herein, a univariate Cox regression analysis technique was utilized to evaluate the influence of RBM4 on disease prognosis, with a specific focus on OS, PFI, and DSS. The sample selection process in this study involved the selection of tumors with a p-value 0.05 and a sample size exceeding 500. A nomogram model was employed for predicting the predict OS in individual patients, which proved to be both effective and convenient. The accuracy of the predictions was validated by constructing 1-, 3-, and 5-year calibration curves.

The TCGA dataset was assessed with the Tumor Immune Estimation Resource 2.0 (https://cistrome.shinyapps.io/timer/) to determine the link between the RBM4 expression levels and the immune cells like B cells, CD4${}^{+}$ T cells, macrophages, and CD8${}^{+}$ T cells. For this purpose, several algorithms like EPIC, CIBERSORT-ABS, CIBERSORT, TIMER, TIDE, XCELL, QUANTISEQ, and MCPCOUNTER were utilized.

To identify the genes that exhibited an RBM4-like expression pattern, a list of the top 100 RBM4-related genes was acquired from the GEPIA2 database. Thereafter, the gene ontology (GO) analysis was carried out, which included biological pathways (BP), cellular components (CC), and molecular functions (MF). These genes were further examined with the Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis to gain further insights into the potential functions of RBM4. For constructing the protein-protein interaction (PPI) network, 100 RBM4-linked genes were used in the Search Tool for the Retrieval of Interacting Genes (STRING) database (https://cn.string-db.org/). The minimum interaction threshold was set at 0.4.

The DESeq R software (Version 4.1.1, R Foundation for Statistical Computing, Vienna, Austria) was used in this study to carry out differential expression analysis of RBM4 in case of cancers that can affect prognosis [23]. Based on the results, we performed GSEA using the clusterProfiler R package (https://bioconductor.org/packages/release/bioc/html/clusterProfiler.html) to study the differential expression of RBM4 in various types of cancer [24].

The HepG2, Huh7, and NCI-H295R cell lines were procured from the Chinese Academy of Sciences and maintained in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% heat-inactivated fetal bovine serum (FBS) sourced from Gemini Bio Products. The cell lines underwent validation through STR profiling and tested negative for mycoplasma. Cultures were incubated at 37 °C in a humidified atmosphere with 5% CO${}_2$. To silence RBM4 expression, lentiviruses were employed, which carried complementary oligonucleotide sequences targeting RBM4. Lentivirus, including complementary oligonucleotide sequences, target sequences were as follows: 5${}^{\prime }$-GAGACAAGCUAGCACUUUAAG-3${}^{\prime }$. These lentiviral constructs were meticulously synthesized by Genomeditech. Following transfection, stably transfected cell lines were selected using 2 µg/mL of puromycin (catalog no. ST551; Beyotime Institute of Biotechnology, Shanghai, China) after a 72-hour incubation period.

For flow cytometry analysis of the proliferating cells, a Cell-Light EdU Apollo 488 In vitro Flow Cytometry Kit (RiboBio, Guangzhou, China) was used to examine EdU-positive cells according to the manufacturer’s protocol.

As for the cells cultured on plates, they were initially fixed with 4% paraformaldehyde and subsequently treated with a 1% Triton solution. Following this, both the sections and cells underwent a blocking step using 10% serum for 30 minutes at room temperature. Next, primary antibodies were applied overnight at 4 °C. The following day, the sections or cells were thoroughly washed and then exposed to secondary antibodies labeled with fluorophores (Thermo Fisher, A48282 and A48287, at a dilution of 1:500, Notch1 Rabbit pAb (A7636), Cyclin D1 Rabbit mAb (A19038), Phospho-mTOR-S2448 Rabbit pAb (AP0094)) for 1 hour at room temperature. Finally, an antifade solution containing 4${}^{\prime }$,6-diamidino-2-phenylindole (DAPI) was introduced, and the resulting fluorescence signals were captured utilizing a fluorescence microscope.

HepG2 and Huh7 cells were initially placed into 6-well plates. After 24 hours, the cells were either transfected with siNC, siRBM4, or exposed to a control adenovirus. Forty-eight hours later, the cells had reached full confluence, and a cell monolayer was gently scraped using a 200-µL pipette tip. The well was rinsed twice with serum-free medium (a mixture of 50% MEM and 50% F12) and then refreshed with a fresh batch of serum-free medium. At the beginning (0 hour) and subsequently every 12 hours during incubation, images were taken using a light microscope (LeicaCTR6000 microscope system) at a magnification of $\times$50. These images were then quantitatively analyzed using ImageJ (X64, v. 2.1.4, LOCI, University of Wisconsin, Madison, WI, USA). The cell cycle was analyzed by flow cytometry using a cell cycle analysis kit (C1052, Beyotime).

This study includes 32 patients with Liver cancer who were diagnosed and treated at Shanghai East Hospital Affiliated to Tongji University were collected, from September 2019 to 2023. Hepatocellular carcinoma and matched edge or adjacent tissues combined microarray chip, including 32 cases of hepatocellular carcinoma and matched edge or adjacent liver tissues, one case of cancer at 2 points, one case of edge or adjacent cancer at 1 point.

The R (ver. 4.0.2) software was used for statistically analyzing the data. The Wilcoxon rank-sum test was utilized for comparing the difference between both groups, while the Spearman rank test was utilized to calculate the link between these groups. Furthermore, the univariate and multivariate Cox proportional hazard regression analyses were conducted for screening the factors affecting disease prognosis. The KM analysis with the log-rank test was utilized for carrying out survival analysis. A p-value 0.05 was deemed statistically significant, with *p 0.05, **p 0.01, ***p 0.001, and ****p 0.0001 implying the level of significance.

To investigate the RBM4 expression in pan-cancer, we evaluated the TCGA_GTEx data derived from UCSC. The findings of this analysis suggest that RBM4 exhibited varying expression patterns across a significant number of cancer types. Specifically, RBM4 expression was found to be elevated in ACC, Bladder Urothelial Carcinoma (BLCA), Breast invasive carcinoma (BRCA), Cervical squamous cell carcinoma and endocervical adenocarcinoma (CESC), Cholangiocarcinoma (CHOL), Colon adenocarcinoma (COAD), Lymphoid Neoplasm Diffuse Large B-cell Lymphoma (DLBC), Glioblastoma multiforme (GBM), Head and Neck squamous cell carcinoma (HNSC), Brain Lower Grade Glioma (LGG), Liver hepatocellular carcinoma (LIHC), Lung adenocarcinoma (LUAD), and Lung squamous cell carcinoma (LUSC), while reduced RBM4 expression levels were observed in Esophageal carcinoma (ESCA), Kidney Chromophobe (KICH), Kidney renal clear cell carcinoma (KIRC), Kidney renal papillary cell carcinoma (KIRP), and Ovarian serous cystadenocarcinoma (OV). These outcomes align with the data obtained from the TCGA database (Fig. 1A,B). In addition, the RBM4 expression levels in 18 different types of tumors were analyzed using paired samples from the TCGA dataset (Fig. 1C). Furthermore, the RBM4 protein expression levels in the healthy control and cancer tissue samples derived from different human organs were analyzed using HPA. Furthermore, we extracted representative immunohistochemical (IHC) images from the bladder, lung, liver, pancreas, and stomach tissues to further examine RBM4 expression (Fig. 2). We found that in glioma (GBMLGG), colon adenocarcinoma (COAD), LIHC, ACC and testicular germ cell tumours (TGCT), the expression level of RBM4 in cancer tissues was significantly higher than that in normal control tissues, especially in liver cancer.

Fig. 1.

The RBM4 mRNA expression in pan-cancer. (A) The RBM4 mRNA expression levels in 33 types of tumors in TCGA_GTEx samples. (B) The RBM4 mRNA expression levels in 33 types of tumors in the TCGA database. (C) RBM4 expression in paired samples of 18 types of cancers in TCGA database. (no significant (ns), p $>$ 0.05; *p 0.05; **p 0.01; ***p 0.001). RBM4, RNA-binding motif 4; TCGA, The Cancer Genome Atlas; TPM, transcripts per million; GTEx, Genotype-Tissue Expression; ACC, adrenocortical carcinoma; BLCA, bladder urothelial carcinoma; BRCA, breast invasive carcinoma; CESC, cervical squamous cell carcinoma and endocervical adenocarcinoma; CHOL, cholangiocarcinoma; COAD, colon adenocarcinoma; DLBC, lymphoid neoplasm diffuse large B-cell lymphoma; ESCA, oesophageal carcinoma; GBM, glioblastoma multiforme; HNSC, head-neck squamous cell carcinoma; KICH, kidney chromophobe; KIRC, kidney renal clear cell carcinoma; LAML, acute myeloid leukaemia; LGG, lower grade gLioma; LIHC, liver hepatocellular carcinoma; LUAD, lung adenocarcinoma; LUSC, lung squamous cell carcinoma; MESO, mesothelioma; OV, ovarian serous cystadenocarcinoma; PAAD, pancreatic adenocarcinoma; READ, rectum adenocarcinoma; SARC, sarcoma; SKCM, skin cutaneous melanoma; STAD, stomach adenocarcinoma; TGCT, testicular germ cell tumours; THCA, thyroid carcinoma; THYM, thymoma; UCEC, uterine corpus endometrial carcinoma; UCS, uterine carcinosarcoma; UVM, uveal melanoma.

Fig. 2.

IHC images of RBM4 in the healthy and malignant tissue samples extracted from HPA. GBMLGG, glioma; IHC, immunohistochemical; HPA, Human Protein Atlas. Image available from v15.proteinatlas.org.

To assess the prognostic value of RBM4 in pan-cancer, we performed a KM survival analysis to understand the relationship between RBM4 expression levels and clinical outcomes. Initially, the link between RBM4 expression levels and the OS rate in 33 types of cancers was determined (Fig. 3A), and the findings of this analysis showed that the abnormal RBM4 expression was associated with OS in ACC (Fig. 3B), KIRC (Fig. 3C), LUSC (Fig. 3D) and pancreatic adenocarcinoma (PAAD) (Fig. 3E). RBM4 expression upregulation was related to a low OS rate in ACC and LUSC. Finally, elevated RBM4 levels in PAAD and KIRC indicated a long OS. We also determine the correlation between the RBM4 expression levels and DSS (Fig. 4A), and the findings implied that the expression of RBM4 was related to the DSS of ACC (Fig. 4B), KICH (Fig. 4C), STES (Fig. 4D), and LIHC (Fig. 4E). Moreover, the RBM4 expression showed varying associations with DSS based on the type of cancer. Specifically, in ACC and KICH, upregulation of RBM4 was linked to worse DSS. Additionally, when the link between RBM4 expression levels and PFI was investigated (Fig. 5A), the findings implied that RBM4 expression was linked to PFI in ACC (Fig. 5B), LIHC (Fig. 5C), pheochromocytoma and paraganglioma (PCPG) (Fig. 5D), KIRC (Fig. 5E). The results revealed that a high RBM4 expression represented a worse PFI in ACC, LIHC, and PCPG, but a better PFI in KIRC. Furthermore, ROC curves were presented for the 6 types of tumors, wherein RBM4 expression was related to disease prognosis. These curves demonstrate the diagnostic ability of RBM4 in the above tumors (Supplementary Fig. 1A–F).

Fig. 3.

Correlation between RBM4 expression and OS in pan-cancer analysis. (A) Forest plots presenting the effects of RBM4 expression levels on OS in pan-cancer. (B–E) Effects of RBM4 expression on OS in different cancers like ACC, KICH, LAML, and LIHC, respectively. OS, Overall Survival.

Fig. 4.

Correlation between RBM4 expression and DSS in pan-cancer. (A) Forest Plot depicting the relationship between RBM4 expression levels on DSS in different types of cancers. (B–E) RBM4 survival curves on DSS in ACC, KICH, STES, and LIHC, respectively. DSS, Disease Specific Survival; STES, Stomach and Esophageal carcinoma.

Fig. 5.

Effect of RBM4 expression levels on PFI in pan-cancer. (A) A forest map describing the relationship between RBM4 expression levels on PFI in TCGA. (B–E) The effect of RBM4 expression levels on PFI in ACC, LIHC, PCPG, and KIRC, respectively. PFI, Progression Free Interval; PCPG, Pheochromocytoma and Paraganglioma.

RBM4 expression levels differed significantly in the I, II, III, and IV pathological stages in several types of cancers such as pan-kidney cohort (KICH + KIRC + KIRP) (KIPAN), KIRC, LIHC, and thyroid carcinoma (THCA) (Fig. 6A). The RBM4 expression varied significantly among the GBMLGG, LGG, KIPAN, KIRC, and LIHC patients, depending on their pathological stage (G1/G2/G3/G4) (Fig. 6B). Based on the M0/M1 pathological stage, we observed significant variations in the RBM4 expression levels in TCGT and Uveal Melanoma (UVM) patients (Fig. 6C). The RBM4 expression varied significantly among prostate adenocarcinoma (PRAD), KIRC, and THCA patients depending on their pathological N stage (N0, N1, N2, or N3) (Fig. 6D). Significant differences were observed in the RBM4 expression levels in the KIPAN, PRAD, KIRC, LIHC, THCA, and UVM patients, based on their pathological T1/T2/T3/T4 stage (Fig. 6E).

Fig. 6.

Correlation between the RBM4 gene expression and different clinicopathological stages in pan-cancer. (A) Violin plots showing differential RBM4 expression levels in the various I/II/III/IV pathological stage. (B) Violin plots depicting the differential RBM4 expression levels in the G1/G2/G3 pathological stage. (C) Violin plots highlighting the differential RBM4 expression levels in the M0/M1 pathological stage. (D) Violin plots indicate the differential RBM4 expression levels in the N0/N1/N2/N3 pathological stages. (E) Violin plots showing the differential RBM4 expression levels in T1/T2/T3/T4 pathological stage. (*p 0.05; **p 0.01; ***p 0.001; ****p 0.0001)

Three types of cancers including KIRC, LIHC, and COAD with a sample size $>$500 were used to design separate nomogram models for validating their respective prognostic values. Calibration curves were employed for determining the prediction accuracy of every nomogram model at 1-, 3-, and 5-year intervals. The findings indicated that RBM4 was involved in the prognosis and exhibited good predictive performance regarding OS for KIRC (Fig. 7A), LIHC (Fig. 7C), and COAD (Fig. 7E) in the nomogram models. Moreover, the calibration results indicated that the developed nomogram models displayed a higher accuracy in predicting the OS, as evidenced by the 1-, 3-, and 5-year OS prediction curves (Fig. 7B,D,F).

Fig. 7.

Nomogram models were constructed and assessed in KIRC, LIHC, and LUSC. (A) Constructing a nomogram model that involves RBM4 expression in KIRC. (B) Calibration curves were employed to assess the nomogram model in KIRC at 1-year, 3-year, and 5-year. (C) Designing a nomogram model with RBM4 expression in LIHC. (D) The 1-year, 3-year, and 5-year calibration curves were utilized for assessing the prediction accuracy of the nomogram model in LIHC. (E) Developing a nomogram model with RBM4 expression in COAD. (F) To assess the prediction accuracy of the nomogram model in COAD, the 1-, 3-, and 5-year calibration curves were utilized.

Immune cells have a crucial role in preserving the body’s immune equilibrium and defending against external threats. Over recent years, a mounting body of research has highlighted the substantial importance of RNA-binding proteins in modulating the expression and regulation of immune cells within the immune response regulation [25, 26]. In this study, we analyzed immune infiltration levels using three established algorithms (xCELL, EPIC, and TIMER) to assess the link between immune cells and RBM4 expression levels across different cancers. The results implied that the RBM4 expression significantly promoted immune cell infiltration, particularly in natural killer cells, macrophages, and neutrophils. The deconvo_xCell method was used to generate 67 immune cell infiltration scores for 1080 malignant samples across 44 types of cancers. To determine the association between genes and the immune cell infiltration scores in every type of cancer, the corr.test function from the psych tool in R was used to estimate Pearson’s correlation coefficient. Through this analysis, we identified a significant link between immune infiltration and RBM4 expression levels in the 44 types of cancers that were studied (Fig. 8A). The Timer method was used to acquire 6 distinctive immune cell infiltration scores, specifically for macrophages, CD4${}^{+}$ T cells, B cells, neutrophils, CD8${}^{+}$ T cells, and dendritic cell (DC). The Pearson’s correlation coefficient of immune cell infiltration scores was assessed and the findings revealed that the RBM4 expression was significantly linked with the immune infiltration levels in 33 different types of cancer. In various types of cancers, the RBM4 expression was found to be most positively related to Th1 cells, NK cells, Th2 cells, and CD4${}^{+}$ T cells. Moreover, RBM4 was significantly and positively related to B cell levels in LIHC, PCPG, KIRC, colon adenocarcinoma/rectum adenocarcinoma oesophageal carcinoma (COADREAD), KIPAN, COAD, KICH, and ACC (Fig. 8B). The deconvo_EPIC method was utilized to obtain immune cell infiltration scores for 10,180 malignant samples across 44 types of cancers. The corr.test function in the psych tool of the R software was utilized to determine the correlation coefficients between the immune cell infiltration scores and gene expression levels in each tumor sample were estimated. The findings implied that the RBM4 expression levels were linked to the immune infiltration levels in 36 cancers (Fig. 8C). After analyzing the results of the 3 algorithms, it was noted that RBM4 was closely related to different immune cells in cancers such as LUSC, LIHC, LUAD, KIPAN, PRAD, PAAD, OV, etc. This relationship could be attributed to a significantly positive or negative correlation, which suggests that additional research needs to be conducted to assess the immune cell infiltration levels in the microenvironment of the aforementioned types of cancers. The results of this study offer significant insights into the potential impact of RBM4 on immune responses in cancer, highlighting its potential role in modulating such responses.

Fig. 8.

Correlations between RBM4 expression and tumor immune infiltration. (A) The RBM4 expression is significantly related to the infiltration levels of different immune cells based on xCELL. (B) The RBM4 expression was seen to be significantly related to the infiltration levels of different immune cells in the Timer database. (C) The RBM4 expression was significantly related to the infiltration levels of different immune cells in the EPIC database. DC, dendritic cell; EPIC, Estimating the Proportion of Immune and Cancer cells. (*p 0.05; **p 0.01; ***p 0.001; **** p 0.0001)

The fundamental concept behind single-cell sequencing technology involves the isolation of cells on an individual basis, followed by the sequencing of RNA or DNA from each isolated cell. This methodology demands precise cell sorting and the utilization of high-throughput sequencing techniques. In recent years, single-cell RNA sequencing (scRNA-seq) has emerged as the predominant approach in single-cell sequencing technology [27, 28]. Hence, in order to delve deeper into the molecular mechanisms of the RBM4 gene in tumorigenesis, we explored its impact on various types of tumors at the single-cell level. The results of this analysis highlighted a positive relationship between RBM4 expression and hypoxia, angiogenesis, inflammation, and proliferation. Conversely, RBM4 expression was seen to be negatively associated with DNA damage, DNA repair, cell cycle, invasion, and quiescence (Fig. 9A). Herein, the association between RBM4 and functional status in different types of cancer was evaluated. Our findings revealed a positive correlation between RBM4 and various cancer-related processes such as proliferation, metastasis, differentiation, inflammation, quiescence, Epithelial-Mesenchymal Transition (EMT), angiogenesis, hypoxia, and apoptosis in Acute Myeloid Leukemia (AML). Additionally, RBM4 was found to be associated with DNA repair mechanisms in acute lymphoblastic leukemia, angiogenesis in Pheochromocytoma (PC), and angiogenesis, differentiation, and inflammation in RetinoBlastoma (RB). The RBM4 gene expression showed a negative correlation with DNA repair in AML and RB, DNA damage and cell cycle in RB, invasion in Uveal Melanoma (UM) and Ovarian serous cystadenocarcinoma (OV), and EMT in PC. Additionally, RBM4 was also seen to be negatively associated with apoptosis and metastasis in AML (Fig. 9B–F). To determine the biological function of RBM4 in different tumors, 100 RBM4-related genes were acquired from the GEPIA2 database. The findings of the GO analysis (Fig. 10A) revealed that RBM4-related genes were involved in “organelle fission”, “mitotic nuclear division”, “nuclear division”, “mitotic sister chromatid segregation”, and other biological processes. These genes could play a role in “chromosomal region”, “spindle”, and other cellular components. It is involved in “microtubule binding”, “tubulin binding”, “microtubule motor activity”, “chemokine activity”, and other molecular functions. Furthermore, the findings of the KEGG pathway analysis (Fig. 10B) implied that RBM4-linked genes may be linked to “Cell cycle”, “Cellular senescence”, “Oocyte meiosis”, “cytokine and cytokine receptor”, “Viral protein interaction”, and “PPAR signaling pathway”. To determine the function of RBM4, we carried out a GSEA depending on the differential expression analysis of RBM4. This helped in elucidating the biological function of RBM4 in three tumors, namely ACC, LIHC, and BRCA, wherein the RBM4 expression was related to the disease prognosis as presented in Fig. 10C. The findings indicate that RBM4 is mainly associated with DNA Replication, Mitotic G1 Phase and G1S Transition, and Cell Cycle Checkpoints.

Fig. 9.

Relationship between RBM4 and functional state in different types of cancers. (A) The interactive bubble chart depicts the correlation between the RBM4 and functional state in 16 cancers. The correlation between the RBM4 and functional state in (B) Acute Myeloid Leukemia (AML), (C) Uveal Melanoma (UM), (D) Pheochromocytoma (PC), (E) RetinoBlastoma (RB), and (F) Ovarian serous cystadenocarcinoma (OV). X-axis depicts different gene sets; ***p 0.001.

Fig. 10.

Functional enrichment analysis of RBM4-linked genes. (A) GO enrichment analysis of 100 RBM4-related genes, which includes BP, CC, and MF. (B) KEGG pathways analysis using the 100 RBM4-related genes. (C) GSEA based on differential expression analysis in ACC, LIHC, and BRCA, respectively. GO, gene ontology; KEGG, Kyoto Encyclopedia of Genes and Genomes.

To further investigate the role of RBM4 in liver cancer, we chose to study the liver cancer cell line HepG2 and Huh7. We collected a total of 32 paired samples of hepatocellular carcinoma (HCC) and adjacent tissues, which were then assembled into a tissue microarray for analysis. Immunohistochemistry revealed that the staining intensity of RBM4 in the tumor group was significantly higher than that in the adjacent tissue group (Supplementary Fig. 2A,B). We utilized short hairpin RNA (shRNA) to knock down RBM4, confirming the effectiveness of interference through validation experiments (Supplementary Fig. 2C,E). Following RBM4 knockdown in the HepG2 and Huh7 cell line, we conducted EdU experiments, which showed a notable decrease in the percentage of cells displaying positive EdU signals (Supplementary Fig. 2D,F). This suggests a reduction in DNA replication activity and weakened cell proliferation ability upon RBM4 downregulation. To assess the impact of RBM4 on cell migration, we examined whether silencing RBM4 influenced the migration of HepG2 cells. As anticipated, the results showed impaired migration abilities in cells with downregulated RBM4 (Supplementary Fig. 2G,H). After knockdown of RBM4 in the NCI-H295R cell line of ACC, the EdU assay showed a significant decrease in the percentage of cells with edu-positive signals, indicating that DNA replication activity was reduced and ACC cancer cell proliferation was weakened after downregulation of RBM4 expression (Supplementary Fig. 2I,J). GSEA enrichment analysis showed that genes related to differential expression of RBM4 in ACC were mainly enriched in the cell cycle. Knockdown of RBM4 in the NCI-H295R cell line of ACC can inhibit the cell cycle G1-S phase transposition, thereby inhibiting the proliferation (Supplementary Fig. 2K). After knocking out RBM4, ClinD1, P-mTOR and NOTCH1 proteins were significantly decreased by immunofluorescence, indicating that RBM4 could affect the development of ACC through PI3K-Akt and NOTCH signaling pathways (Supplementary Fig. 2L-N). Together, these findings provide further evidence supporting the role of RBM4 as a potential target LIHC and ACC.